ОПТИМИЗАЦИЯ МНОГОСТУПЕНЧАТОГО ПРОЕКТА НЕФТЯНОГО МЕСТОРОЖДЕНИЯ В УСЛОВИЯХ НЕОПРЕДЕЛЕННОСТИ

Изобретение относится к нефтяной промышленности и может быть использовано для управления операциями нефтяного месторождения в условиях неопределенности. В частности предложен способ управления операциями нефтяного месторождения, включающий: получение модели геологической среды, содержащей модель проекта трещины, имеющей свойство трещины с неопределенным значением; получение набора характерных значений, которые отражают неопределенность в свойстве трещины, посредством получения набора характерных значений, представляющих неопределенность в модели скорости распространения сейсмических волн; получение данных микросейсмического события; генерирование набора характерных значений для свойства трещины посредством использования набора характерных значений для модели скорости распространения сейсмических волн и данных микросейсмического события, решение задачи оптимизации нефтяного месторождения с переменным параметром управления посредством использования набора характерных значений для свойства ...

СПОСОБ И УСТРОЙСТВО ДЛЯ НЕЙРОННОГО ВРЕМЕННОГО КОДИРОВАНИЯ, ОБУЧЕНИЯ И РАСПОЗНАВАНИЯ

Группа изобретений относится к нейронным сетям и может быть использована для нейронного временного кодирования, обучения и распознавания. Техническим результатом является уменьшение сложности кодирования. Способ содержит этапы, на которых: используют относительную задержку, которая задерживает один или более синаптических вводов в нейронную схему посредством временных задержек; применяют динамическую спайковую модель для того, чтобы определять спайковое поведение нейронной схемы на основе взвешенных и задержанных синаптических вводов нейронной схемы; регулируют согласно правилу неконтролируемого обучения весовые коэффициенты, ассоциированные с синаптическими вводами, в зависимости от взаимосвязи по синхронизации образования спайков в нейронных схемах и задержанных синаптических вводов; выбирают допускающий повторное использование синапс, ассоциированный с нейронной схемой, на основании уменьшения весового коэффициента допускающего повторное использование синапса посредством правила неконтролируемого ...

УСОВЕРШЕНСТВОВАНИЯ В НЕЙРОННЫХ СЕТЯХ

Изобретения относятся к вычислительной технике и могут быть использованы для моделирования нейронных сетей. Техническим результатом является расширение класса решаемых задач. Запоминающие устройства предназначены для хранения отклика и соответствующего по крайней мере одного входного аналогового стимула, каждый из которых имеет соответствующее значение из заданных предварительно назначенных комплексных полярных значений, и включают переписываемое корреляционное средство, упомянутый отклик соответствует множеству стимулов. Каждая векторная величина имеет модуль, ограниченный диапазоном вероятностного распределения, и коэффициент фазового угла, представляющий заданные семантически полезные характеристики упомянутых стимулов и соответствующего отклика. 2 с. и 45 з.п. ф-лы, 31 ил.

СПОСОБ, УСТРОЙСТВО И СЕРВЕР ДЛЯ ОПРЕДЕЛЕНИЯ ПЛАНА СЪЕМКИ ИЗОБРАЖЕНИЯ

Изобретение относится к определению плана съемки изображения. Техническим результатом является повышение точности классификации изображений. В способе получают галерею пользовательского терминала; осуществляют идентификацию и маркировку изображения; получают обучающий набор выборки; вводят каждую из множества последовательностей обучающих изображений; обучают коэффициенты признака между уровнями скрытых узлов модели определения исходного изображения плана съемки; получают тестовый набор выборки; осуществляют идентификацию тестовых изображений; определяют точность классификации модели определения плана съемки изображения; если точность классификации меньше заданного порогового значения, выполняют: обновление обучающего набора выборки; обучение, в соответствии с обновленным обучающим набором выборки, коэффициентов признака между соответствующими уровнями скрытых узлов модели определения плана съемки изображения; выполнения итерации обновления модели определения плана съемки изображения; выполнение ...

Специализированная вычислительная система, предназначенная для вывода в глубоких нейронных сетях, основанная на потоковых процессорах

Изобретение относится к вычислительной технике. Техническим результатом является создание вычислительной системы для выполнения нейросетевых алгоритмов. Вычислительная система содержит массив вычислительных ядер, чередующихся с блоками локальной памяти, блок синхронизации, блок диагностики, краевые интерфейсные блоки, вспомогательное процессорное ядро, процессор общего назначения, динамическую оперативную память, контроллер динамической оперативной памяти, контроллер интерфейса, использующийся для подключения к внешней компьютерной системе. 9 з.п. ф-лы, 2 ил.

БАЙЕСОВСКОЕ РАЗРЕЖИВАНИЕ РЕКУРРЕНТНЫХ НЕЙРОННЫХ СЕТЕЙ

Изобретение относится к области искусственного интеллекта, и в частности, к рекуррентным нейронным сетям (РНС). Техническим результатом является повышение степени сжатия. Предложен новый метод байесовского разреживания для рекуррентных архитектур с гейтами, в котором учитываются их рекуррентные особенности и механизм с гейтами. В предложенном методе удаляются нейроны из ассоциированной модели и гейты делаются константными, что обеспечивает не только сжатие сети, но и значительное ускорение прохода вперед. На дискриминативных задачах данный метод обеспечивает максимальное сжатие LSTM, так что остается лишь небольшое количество входных и скрытых нейронов при незначительном снижении качества. Такую малую модель легко интерпретировать. 3 н. и 16 з.п. ф-лы, 2 ил., 2 табл.

УСТРОЙСТВО ОБРАБОТКИ ДАННЫХ, СПОСОБ ОБРАБОТКИ ДАННЫХ И НОСИТЕЛЬ ДАННЫХ

Изобретение относится к вычислительной технике. Технический результат заключается в повышении производительности нейронных сетей. Устройство обработки данных содержит блок обработки данных для обработки входных данных с использованием нейронной сети; блок управления сжатием для определения шагов квантования и генерирования информации квантования, которая определяет шаги квантования, причем шаги квантования используются, когда квантуются данные параметров нейронной сети; и блок кодирования для кодирования информации конфигурации сети и информации квантования, чтобы сгенерировать сжатые данные, причем информация конфигурации сети включает в себя данные параметров, квантованные с использованием шагов квантования, определенных блоком управления сжатием. 4 н. и 4 з.п. ф-лы, 16 ил.

АНСАМБЛЬ КЛАССИФИКАТОРОВ ДЛЯ ОБНАРУЖЕНИЯ АНОРМАЛЬНЫХ СЕРДЕЧНЫХ ТОНОВ

КЛАССИФИКАЦИЯ САЙТА СПЛАЙСИНГА НА ОСНОВЕ ГЛУБОКОГО ОБУЧЕНИЯ

СИСТЕМА ГЕНЕРИРОВАНИЯ И ТЕСТИРОВАНИЯ ВЕБ-ИНТЕРФЕЙСА НА ОСНОВЕ МАШИННОГО ОБУЧЕНИЯ

НАЗНАЧЕНИЕ ОЦЕНКИ РЕЛЕВАНТНОСТИ ДЛЯ ИСКУССТВЕННЫХ НЕЙРОННЫХ СЕТЕЙ

СИСТЕМЫ И СПОСОБЫ ОБУЧЕНИЯ ОБНАРУЖЕНИЮ ЛИСТВЫ

Способ многопризнакого распознавания в многофункциональной радиолокационной станции класса летательного аппарата по принципу "самолет с турбореактивным двигателем - самолет с турбовинтовым двигателем - вертолет - ракета - беспилотный летательный аппарат" на основе совместного применения калмановской фильтрации и нейронной сети

Изобретение относится к области радиолокации и может быть использовано для распознавания в многофункциональной радиолокационной станции (РЛС) класса летательного аппарата (ЛА) по принципу «самолет с турбореактивным двигателем (ТРД) - самолет с турбовинтовым двигателем (ТВД) - вертолет - ракета - беспилотный летательный аппарат (БпЛА)» на основе совместного применения калмановской фильтрации и нейронной сети (НС). Технический результат заключается в создании способа, позволяющего по многопризнаковому пространству распознать в многофункциональной РЛС с вероятностью, не ниже заданной, класс ЛА по принципу «самолет с ТРД - самолет с ТВД - вертолет - ракета - БпЛА». Способ заключается в том, что отраженный от ЛА радиолокационный (РЛ) сигнал принимают многофункциональной РЛС и подвергают на промежуточной частоте узкополосной доплеровской фильтрации на основе процедуры быстрого преобразования Фурье (БПФ), преобразуют в амплитудно-частотный спектр (АЧС), спектральные составляющие которого обусловлены ...

ANORDNUNG ZUR SIGNALVERARBEITUNG

Rechenvorrichtung

Eine Rechenvorrichtung umfasst: eine Rückschlussschaltung, die ein Erkennungsergebnis eines Erkennungsziels und die Zuverlässigkeit des Erkennungsergebnisses unter Verwendung von Sensordaten von einer Sensorgruppe, die das Erkennungsziel detektiert, und eines ersten Klassifikators, der das Erkennungsziel klassifiziert, berechnet; und eine Klassifikationsschaltung, die die Sensordaten entweder in ein zugeordnetes Ziel, dem das Erkennungsergebnis zugeordnet ist, oder ein nicht zugeordnetes Ziel, dem das Erkennungsergebnis nicht zugeordnet ist, auf der Basis der Zuverlässigkeit des Erkennungsergebnisses, die durch die Rückschlussschaltung berechnet wird, klassifiziert.

TRAINIEREN EINES NEURONALEN NETZES ZUM BESTIMMEN VON FUSSGÄNGERN

Die Offenbarung stellt Trainieren eines neuronalen Netzes zum Bestimmen von Fußgängern bereit. Es werden ein Trainingssystem für ein neuronales Netzsystem und ein Trainingsverfahren offenbart. Das Verfahren kann Folgendes umfassen: Empfangen eines Einzelbildes von einem Sensor, das aufgenommen wurde, während ein Fahrzeugführer ein Fahrzeug steuert; Verwenden eines dem Sensor zugeordneten Augennachverfolgungssystems, das die Augen des Fahrzeugführers überwacht, um Augapfelblickdaten zu bestimmen; Bestimmen einer Vielzahl von Fußgängern aus dem Einzelbild; und iteratives Trainieren des neuronalen Netzsystems, um unter Verwendung der Augapfelblickdaten und eines Antwortdatensatzes, der auf den Augapfelblickdaten basiert, den einen oder die mehreren Zielfußgänger aus der Vielzahl von Fußgängern zu bestimmen, wobei der bestimmte eine oder die bestimmten mehreren Zielfußgänger eine relativ höhere Wahrscheinlichkeit einer Kollision mit dem Fahrzeug aufweisen als ein Rest der Vielzahl von Fußgängern ...

Memristive Einheit auf Grundlage einer Alkali-Dotierung von Übergangsmetalloxiden

Eine memristive Einheit beinhaltet eine erste leitfähige Materialschicht. Eine Oxidmaterialschicht ist auf der ersten leitfähigen Schicht angeordnet. Eine zweite leitfähige Materialschicht ist auf der Oxidmaterialschicht angeordnet, wobei die zweite leitfähige Schicht eine Metall-Alkali-Legierung aufweist.

Neuronales Netzwerk

Hybride Kraftfahrzeug-Sensorvorrichtung mit einem neuronalen Netz und einem Bayes'schen Filter, sowie Verfahren zum Betreiben einer solchen Kraftfahrzeug-Sensorvorrichtung

Die Erfindung betrifft ein Verfahren zum Betreiben einer Sensorvorrichtung (2) eines Kraftfahrzeugs (1), mit einem a) Bereitstellen von einer Messinformation durch eine Sensoreinheit (5) der Sensorvorrichtung (2); einem b) Klassifizieren der bereitgestellten Messinformation durch ein neuronales Netz (6) der Sensorvorrichtung (2); und einem c) Filtern eines Ergebnisses des Klassifizierens mittels eines Bayes'schen Filters (7) der Sensorvorrichtung (2), sodass das Ergebnis in Abhängigkeit einer für den Bayes'schen Filter (7) vorgegebenen Zustandsinformation an einen von der Zustandsinformation repräsentierten Zustand des Kraftfahrzeugs (1) oder eines Objekts (4) in der Umgebung (3) des Kraftfahrzeugs (1) angepasst wird, um die Schwächen bekannter Sensorvorrichtungen zu überwinden und das Verarbeiten einer Messinformation zu verbessern.

Generieren einer Ausgabe für eine Ausgabeschicht eines neuronalen Netzwerks

Systeme, Verfahren und Vorrichtungen, die Computerprogramme beinhalten, die auf einem Computerspeichermedium zum Verarbeiten einer Netzwerkeingabe durch ein neuronales Netzwerk codiert sind, das eine oder mehrere anfängliche Netzwerkschichten aufweist, auf die eine Softmax-Ausgabeschicht folgt. In einem Aspekt beinhalten die Verfahren Erhalten einer durch die eine oder mehreren anfänglichen neuronalen Netzwerkschichten generierten Schichtausgabe und Verarbeiten der Schichtausgabe durch die Softmax-Ausgabeschicht, um eine neuronale Netzwerkausgabe zu generieren. Verarbeiten der Schichtausgabe durch die Softmax-Ausgabeschicht beinhaltet Bestimmen einer Anzahl von Vorkommnissen in den Schichtausgabewerten für jeden möglichen Ausgabewert; für jeden in den Schichtausgabewerten vorkommenden Ausgabewert bestimmen eines jeweiligen Potenzierungsmaßes; Bestimmen eines Normalisierungsfaktors für die Schichtausgabe durch Kombinieren der Potenzierungsmaße gemäß der Anzahl von Vorkommnissen der möglichen ...

Abriebgrössen-Schätzvorrichtung und Abriebgrössen-Schätzverfahren für das Rückschlagventil einer Spritzgiessmaschine

Eine Abriebgrößen-Schätzvorrichtung speichert ein durch auf der Basis einer Merkmalsgröße, die von einer physikalischen Größe, die in einem durch eine Spritzgießmaschine durchgeführten Einspritzen erfasst wird, und einer Information in Bezug auf eine Abriebgröße eines Rückschlagventils, das an der Spritzgießmaschine beim Einspritzen befestigt wurde, durchgeführtes überwachtes Lernen ermitteltes Ergebnis. Die Abriebgrößen-Schätzeinheit schätzt eine Abriebgröße eines Rückschlagventils, das an der Spritzgießmaschine beim Einspritzen befestigt wurde, auf der Basis des gespeicherten Lernergebnisses und der extrahierten Merkmalsgröße.

Fusionssystem zur Fusion von Umfeldinformation für ein Kraftfahrzeug

Ein Aspekt der Erfindung betrifft ein Fusionssystem für ein Kraftfahrzeug, wobei das Fusionssystem zumindest zwei Umfeldsensoren, ein mit den Umfeldsensoren gekoppeltes neuronales Netz zur Fusion von Umfeldinformation der Umfeldsensoren, eine Fusionseinrichtung zur Fusion von Umfeldinformation der Umfeldsensoren, eine mit der Fusionseinrichtung dem neuronalen Netz gekoppelte Kontrollvorrichtung umfasst, und die Kontrollvorrichtung eingerichtet ist, die durch das neuronale Netz fusionierte Umfeldinformation in Abhängigkeit von durch die Fusionseinrichtung fusionierte Umfeldinformation anzupassen, und die angepasste Umfeldinformation einem Fahrerassistenzsystem des Kraftfahrzeugs bereitzustellen.

Canonical Spiking Neuron Network for Spatiotemporal Associative Memory

Neural network compute tile

A computing unit has one or more multiply-accumulate units 215 operating in parallel. The units take a first input from a narrow memory 210 and a second input from a wide memory 212. The addresses read from the narrow memory are selected by a traversal unit. The traversal unit may implement nested loops of operations on the data in the narrow memory. It may use a stride value to select entries at a given address interval. A broadcast bus 216 may connect the narrow memory to all of the multiply-accumulate units. The bus may be wider than the values passed to the multiply accumulate units, to allow multiple values to be transported at once. The results may be output with a pipelined shift register 236 to an output bus 218. A non-linear tensor function 224 may be applied to the result of the multiply-accumulate operations. A ring bus 205 may connect the unit to other devices.

Large-scale image tagging using image-to-topic embedding

A method of calculating the relevance of tags applied to images involves receiving an image 314 with associated tags, which may be user-applied tags. The tags are used to create a weighted word vector 218, also known as a soft topic vector, which represents the dominant concept among the keyword tags. Visual features 312 of the image may be used to create an image feature vector 310 which can then be aligned in a common embedding space. The aligned vectors can then be used to calculate a relevancy score for each tag as it pertains to the image. The visual features may be determined with a convolutional neural network. The weighted word and image feature vectors may be aligned using cosine similarity loss. Each tag may be assigned a word vector representation and a weighted average of the word vectors can then be used to generate the weighted word vector.

Facial localisation in images

Non-binary context mixing compressor/decompressor

A technique for non-binary context mixing in a compressor includes generating, by a plurality of context models, model predictions regarding a value of a next symbol to be encoded. A mixer generates a set of final predictions from the model predictions. An arithmetic encoder generates compressed data based on received input symbols and the set of final predictions. The received input symbols belong to an alphabet having a size greater than two and the mixer generates a feature matrix from the model predictions and trains a classifier that generates the set of final predictions.

Methods and systems to determine and optimize reservoir simulator performance in a cloud computing environment

Disclosed are systems and methods for allocating resources for executing a simulation. These include receiving a simulation for execution, calculating an initial runtime of an initial time step of the simulation, determining a total runtime of the simulation based on the initial runtime, selecting a runtime model based on the initial time step, total runtime, or a parameter of the simulation, identifying, based on the selected runtime model, an allocation of a resource providing an increase in runtime speed, allocating the identified resource, and executing the simulation using the allocated resource.

Petroleum reservoir behaviour prediction using a proxy flow model

Using production data and a production flow record based on the production data, a deep neural network (DNN) is trained to model a proxy flow simulation of a reservoir. The proxy flow simulation of the reservoir is performed, using an ensemble Kalman filter (EnKF), based on the trained DNN. The EnKF assimilates new data through updating a current ensemble to obtain history matching by minimizing a difference between a predicted production output from the proxy flow simulation and measured production data from a field. Using the updated current ensemble, a second proxy flow simulation of the reservoir is performed based on the trained DNN. The assimilating and the performing are repeated while new data is available for assimilating. Predicted behavior of the reservoir is determined based on the proxy flow simulation of the reservoir. An indication of the predicted behavior is provided to facilitate production of fluids from the reservoir.

Hardware implementation of a neural network

A neural network, e.g. convolutional-neural-network (CNN), has first layer 502-1 receiving input data 504 and end layer 502-2 generating output data 508. Off-chip SDRAM DDR memory 312 stores input data, weights, and output data. Generated output data comprises p data-set planes 508 dependent on weights, each plane comprising n>2 output data elements. Processing of inputs proceeds depth‑wise through each layer by reading a first block of input data and weights, then evaluating each layer to calculate a first block of output data and writing it to memory 312. The first output block has multiple data elements m, with m>1, m Подробнее

Joint shape and texture decoders for three-dimensional rendering

A method of training a mesh decoder neural network 112 for three-dimensional (3D) image modeling, for generating a shape and texture map 114 of input visual data 102, 102a, 102b is described. The training comprises generating embedding parameters 106 of input images, representing the geometry and texture of the input data. The decoder neural network is applied to the embedding parameters, generating a shape and texture map of the input, and includes one or more geometric convolutional layers (502, Fig. 5). The parameters of the mesh decoder are updated, comparing the input data with output pictures derived from the generated shape and texture map. Input data may include picture and/or shape and texture maps, and might describe face images. The decoder may be structured with alternating upscaling (606, Fig. 6) and geometric convolutional neural network layers. A mesh encoder 104 may be used to generate the embedded parameters and may include a number of downsampling, or pooling, layers ( ...

Closed loop automatic dataset creation systems and methods

Various techniques are provided for training a neural network to classify images. A convolutional neural network (CNN) is trained using training dataset comprising a plurality of synthetic images. The CNN training process tracks image-related metrics and other informative metrics as the training dataset is processed. The trained inference CNN may then be tested using a validation dataset of real images to generate performance results (e.g., whether a training image was properly or improperly labeled by the trained inference CNN). In one or more embodiments, a training dataset and analysis engine extracts and analyzes the informative metrics and performance results, generates parameters for a modified training dataset to improve CNN performance, and generates corresponding instructions to a synthetic image generator to generate a new training dataset. The process repeats in an iterative fashion to build a final training dataset for use in training an inference CNN.

Resistive processing unit architecture with separate weight update and inference circuitry

Systems and methods are provided to perform weight update operations in a resistive processing unit (RPU) system to update weight values of RPU devices comprising tunable resistive device. A weight update operation for a given RPU device includes maintaining a weight update accumulation value for the RPU device, adjusting the weight update accumulation value by one unit update value in response to a detected coincidence of stochastic bits streams of input vectors applied on an update row and update column control lines connected to the RPU device, generating a weight update control signal in response to the accumulated weight value reaching a predefined threshold value, and adjusting a conductance level of the tunable resistive device by one unit conductance value in response to the weight update control signal, wherein the one unit conductance value corresponds to one unit weight value of the RPU device.

APPARATUS FOR PATTERN RECOGNITION

Business discovery from imagery

Convolutional neural network hardware configuration

A dialogue system, a dialogue method and a method of adapting a dialogue system

Weight buffers

Learning based Bayesian optimization for optimizing controllable drilling parameters

A method for optimizing real time drilling with learning uses a multi-layer Deep Neural Network (DNN) built from input drilling data. A plurality of drilling parameter features is extracted using the DNN. A linear regression model is built based on the extracted plurality of drilling parameter features. The linear regression model is applied to predict one or more drilling parameters.

Automatic classification of drilling reports with deep natural language processing

An automatic method of generating decision cubes from cross dependent data sets

Methods and systems for training a machine learning model

Method, device and system for adaptive training of machine learning models via detected in-field contextual sensor events and associated located and retrieved

Hierarchical mantissa bit length selection for hardware implementation of deep neural network

Deep learning based reservoir modeling

Recist

Object identification system and method

Low precision efficient multiplication free convolutional filter bank device

A convolutional neural network CNN comprises banks of convolutional filters and input, output, and filter coefficients represented with low-precision significands, preferably 3 or 4 bits, for which no loss of accuracy is found This allows a simple look-up table to replace multiplications in convolutional CNNs for all possible product values of significands of input tensors and filter coefficients. Therefore, the accumulated result across coefficients of each filter is efficiently formed by summing the shifted and filter centre aligned output of this look up table. Implementing convolutional filtering is simplified with less computational cost than devices employing higher-precision and multiplication. Implementations comprise creating and sharing low-precision and padded significand product intermediate results 37, which are indexed by the filter depth index and shifted to from a filter center aligned 2D tensor 26. Tensor 26 is accumulated and reformatted to higher precision (11). Input ...

Video processing

A method comprises: obtaining or receiving video data; providing a current frame and/or one or more previous frames of the obtained or received video data to an input of a neural network (NN); generating a predicted output at an output of the neural network, comprising at least one of one or more predicted future frames of the video data and predicted properties of one or more future frames of the video data; determining one or more processing decisions based, at least in part, on the predicted output; and processing the current frame of the video data at least partially according to the one or more processing decisions. Predicted future frames may encoded for transmission. Processing the current frame may comprise generating residual information based on a difference between a current frame and an earlier prediction of the current frame. The processing decisions may include determining whether to store the current video frame as a reference frame or determining an encoding method for at ...

Performing fernel striding in hardware

Liquid flow distribution using one or more neural networks

A liquid cooling system for a data center comprises a processor with one or more neural networks which can adjust one or more flow control valves 308, 310, to control a variation in liquid flow rate across the components 302 of the data center. The liquid cooling system may include a liquid cooling loop with a reverse return portion (422, Figure 4B) which provides similar path lengths for a plurality of components. The neural network(s) may maintain the variation on liquid flow rate across the data center below a maximum variation threshold. Individual control valves may control the flow of liquid into or out of individual components, and the components may be at least one of server racks, servers, computer components, or cooling plates. The input to the neural network(s) may comprise sensor data including at least one of flow rate, pressure, temperature, fluid velocity, power consumption, or workload for one or more locations in the data center.

Water Non-Water Segmentation Systems And Methods

A method of producing a navigation (range) chart comprising: receiving an image 3440; segmenting the image into water and non-water pixels; and generating a range chart corresponding to the environment about the mobile structure 3450. Generation of the range chart may be performed by a convolutional autoencoder or a self-supervised neural network 3420. The BEV network 3420 may be trained using a semantic segmentation network that is trained on labelled bird eye view images. The BEV network: receives a horizon stabilised visible spectrum image and infrared image from mounted imagers; fuses the visible spectrum and infrared images; creates an autoencoded birds eye view (BEV) image from the fused image; and segments the BEV autoencoded image into water and non-water features using a graph cut image segmenter. The navigation chart may comprise a range of each pixel from the imager, displayed through a range contour (2922, Fig.29). The range (navigation) chart may be augmented with: information ...

Intelligent chat channel processor

An Intelligent Chat Channel Processor (ICCP) 410 for chat channel communications is configured to receive a user message and classify the message into one of a question, an answer, and a statement. When the input is a question, the ICCP determines a set of relevant answers from a database that are related to the question. The ICCP determines a relatedness score reflecting a degree of relatedness between each related answer of the set and the question. A top answer is determined from the set of related answers based on the relatedness scores, and the ICCP presents at least one of the top answer and the set of related answers to the user. When the input is an answer, the ICCP stores the answer in the database, and when the input is classified as a statement, it is discarded. The ICCP receives a feedback user message that rates the top answer and modifies a reward score for it to adjust a future top answer response. The set of relevant answers may be determined using a term frequency-inverse ...

API for recurrent neural networks

Apparatuses, systems, and techniques to implement a recurrent neural network. In at least one embodiment, an application programming interface receives one or more API calls comprising a graph definition and a recurrence attribute, and executes a recurrent neural network based on the graph definition.

Synthetic system fault generation

Systems, computer-implemented methods, and computer program products that facilitate synthetic system fault generation are provided. According to an embodiment, a system comprises a processor that executes the following computer-executable components stored in a non-transitory computer readable medium: a generator component that employs a trained artificial intelligence (AI) model to generate a synthetic system fault, represented as a combination of discrete parameters and continuous parameters that define a system state; and a fault assembler component that analyses the synthetic system fault and generates textual content corresponding to the synthetic system fault.

System and method for customized reminders

A system 100 and method of notifying a user 102 of a meeting comprises accessing a meeting on a user’s calendar, the meeting maintained as a data record in the data storage 110, and accessing a desired start time of the meeting for the user, which may or may not be the start time of the meeting. A number of delay factors are determined corresponding to a number of attributes associated with the meeting that will be encountered by the user between a notification message being presented by a user device and the user joining the meeting. These can include location, network speed, personal needs and resources required. An advance notification time is determined for the meeting comprising a number of delay factors before the desired start time and an advance notification 112 is present on a user’s device 104 at the advance time. A neural network may be trained to determine the delay factors based on past data.

Mixed-precision deep neural network ensemble

Real-time abnormal diagnosis and interpolation method for water regimen monitoring data

A real-time abnormal diagnosis and interpolation method for water regimen monitoring data, which relates to the technical field of water regimen monitoring. The method comprises the following steps: acquiring water regimen monitoring data, drawing a box plot, and performing the real-time identification and diagnosis of abnormal data on the basis of the box plot, performing grey correlation analysis on other variables related to predictor variables; constructing and training a BP neural network model, applying the BP neural network model to predict the water regimen monitoring data in real time, and performing abnormal diagnosis and data interpolation. The use of the described method can effectively improve the real-time prediction and monitoring of the water regimen monitoring data, and may promptly diagnose and interpolate abnormal data, which may thus improve the reliability of data, objectively reflect water regimen changes, and effectively guide project scheduling.

Graphics processing units for detection of cheating using neural networks

Disclosed is a system to detect cheating in a computer game by analysing images 400 generated by the computer game for illicit or abnormal information 402-410. Neural networks are trained to recognise illicit or abnormal information in the images to detect cheating by one or more users of a computer game.

Knowledge discovery using a neural network

A processor comprises one or more circuits to identify one or more relationships among one or more words using one or more transformer-based language neural networks 602 trained using domain-specific data 601. The neural network may have a Bidirectional Encoder Representations from Transformers (BERT) layer and be trained using a Robustly Optimised BERT Approach (RoBERTa). The one or more words may comprise a query phrase 604 and a target phrase 606, and a scoring function may calculate an associated score between each query word and each target word. Medical documents such as clinical trial data can be analysed using the processor to identify drugs having a target property such as efficacy, and drugs may be ranked according to calculated efficacy scores.

Performing non-maximum suppression in parallel

The application describes apparatuses, systems, and techniques to perform non-maximum suppression (NMS) in parallel to remove redundant bounding boxes 606. In at least one embodiment, two or more parallel circuits to perform two or more portions of an NMS algorithm in parallel to remove one or more redundant bounding boxes 606 corresponding to one or more objects within one or more digital images 601.

Object class inpainting in digital images utilizing class-specific inpainting neural networks

The present disclosure relates to systems, methods, and computer readable media that generate inpainted digital images utilizing class-specific cascaded modulation inpainting neural networks. For example, the disclosed systems utilize a class-specific cascaded modulation inpainting neural network that includes cascaded modulation decoder layers to generate replacement pixels portraying a particular target object class. To illustrate, in response to user selection of a replacement region and target object class, the disclosed systems utilize a class-specific cascaded modulation inpainting neural network corresponding to the target object class to generate an inpainted digital image that portrays an instance of the target object class within the replacement region. Moreover, in one or more embodiments the disclosed systems train class-specific cascaded modulation inpainting neural networks corresponding to a variety of target object classes, such as a sky object class, a water object class ...

Real time generative audio for brush and canvas interaction in digital drawing

Embodiments are disclosed for real time generative audio for brush and canvas interaction in digital drawing. The method includes: receiving a user input and a selection of a tool and canvas (e.g. chalk and chalkboard) for generating audio for a digital drawing interaction; generating intermediary audio data based on the user input and the tool selection, wherein the intermediary audio data includes a pitch and a frequency; processing, by a trained audio transformation model and through a series of one or more layers of the trained audio transformation model, the intermediary audio data; adjusting the series of one or more layers of the trained audio transformation model to include one or more additional layers to produce an adjusted audio transformation model and generating, by the adjusted audio transformation model, an audio sample based on the intermediary audio data. The audio data provides a realistic sound corresponding to the selected tool and canvas for an enhanced user experience ...

Methods for training and analysing input data using a machine learning model

Broadly speaking, the present techniques generally relate to machine learning models comprising neural network layers, in which the quantisation level of each layer of the model can be independently selected at run-time. In particular, the present application relates to a computer-implemented method for analysing input data on a device using a trained machine learning, ML, model, comprising independently selecting a quantisation level for each of a plurality of network layers of the model at runtime. The present application also relates to a computer-implemented method of training a machine learning model so that the quantisation level of each of the plurality of network layers is independently selectable at runtime. A single trained model with a single set of weights can therefore be deployed, with the quantisation of each layer selected at runtime to suit the capabilities of the device and available resource.

Techniques for identification of out-of-distribution input data in neural networks

Apparatuses, systems, and techniques to identify out-of-distribution input data in one or more neural networks. In at least one embodiment, a technique includes training one or more neural networks to infer a plurality of characteristics about input information based, at least in part, on the one or more neural networks being independently trained to infer each of the plurality of characteristics about the input information.

NEURAL NETWORK, MECHANISM FOR PROCESSING INFORMATION, PROCEDURES FOR THE OPERATION OF A NEURAL NETWORK, A PROGRAM ITEM AND A COMPUTER-READABLE MEDIUM

CATEGORIZATION AUTOMATISM, WHICH USES NEURAL GROUP SELECTION WITH RE-ENTRY

Efficient data layouts for convolutional neural networks

Systems and methods for efficient implementation of a convolutional layer of a convolutional neural network are disclosed. In one aspect, weight values of kernels in a kernel stack of a convolutional layer can be reordered into a tile layout with tiles of runnels. Pixel values of input activation maps of the convolutional layer can be reordered into an interleaved layout comprising a plurality of clusters of input activation map pixels. The output activation maps can be determined using the clusters of the input activation map pixels and kernels tile by tile.

End-to-end speaker recognition using deep neural network

The present invention is directed to a deep neural network (DNN) having a triplet network architecture, which is suitable to perform speaker recognition. In particular, the DNN includes three feed-forward neural networks, which are trained according to a batch process utilizing a cohort set of negative training samples. After each batch of training samples is processed, the DNN may be trained according to a loss function, e.g., utilizing a cosine measure of similarity between respective samples, along with positive and negative margins, to provide a robust representation of voiceprints.

Quasi-recurrent neural network

The technology disclosed provides a quasi-recurrent neural network (QRNN) that alternates convolutional layers, which apply in parallel across timesteps, and minimalist recurrent pooling layers that apply in parallel across feature dimensions.

IMAGE COMPLETION WITH IMPROVED DEEP NEURAL NETWORKS

Digital image completion using deep learning is described. Initially, a digital image having at least one hole is received. This holey digital image is provided as input to an image completer formed with a framework that combines generative and discriminative neural networks based on learning architecture of the generative adversarial networks. From the holey digital image, the generative neural network generates a filled digital image having hole-filling content in place of holes. The discriminative neural networks detect whether the filled digital image and the hole filling digital content correspond to or include computer-generated content or are photo realistic. The generating and detecting are iteratively continued until the discriminative neural networks fail to detect computer-generated content for the filled digital image and hole-filling content or until detection surpasses a threshold difficulty. Responsive to this, the image completer outputs the filled digital image with hole-filling ...

A Sentiment Analysis System Based on Deep Learning

Abstract This application focuses on a sentiment analysis system based on deep learning, a process which consists of analysis, processing, classification and induction to subjective texts with emotional prone. Firstly, it conducts some imperative preparation for the emotional text such as cleaning up all html link, symbol and punctuation before we divide sentences into words and then turn these words into words list in order to help sorting different emotion data by using Natural Language Toolkit (NTLK) model. After the preprocessing, the data is already transferred to a network, whose framework and operation is carefully devised. Then, the network is fed with enough training data to train and adjust the model which just trained. Now the network is tested with random testing data and evaluated by its prediction accuracy. This application can show its value in a broad scope of sentiment analysis such as product evaluation and film review. It is also a remarkable application making it convenient ...

Ad hoc neural network for proof of wallet

A proof of wallet approach is used for transaction validation for a digital currency. When a transaction is requested, a set of witness nodes are selected to form an ad hoc neural network. The witness nodes may be client devices of other users of the digital currency. Each witness node receives input information about the transaction (e.g., an encrypted amount and nonce) and neural network parameters (e.g., input weights and a bias). The input information passes through the ad hoc neural network, which generates an output validation value. The transaction is approved if the output validation value is consistent with a verification value generated from the transaction parameters, neural network parameters, and digital currency information stored on a blockchain. If the transaction is approved, the transaction is added to the blockchain in conjunction with the identity of the witness nodes and any other pertinent information Client Device Client Device BClient Device ---- 11ON Network Transaction ...

Radiation therapy planning using deep convolutional network

A deep convolutional neural network can be trained to provide a patient radiation treatment plan. Training can include collecting patient data based on at least one image of patient anatomy from patients, determining a treatment plan including a set of control points from the collected patient data, and using the determined treatment plans and the corresponding collected patient data to train a deep convolutional neural network for regression to determine a treatment plan including a set of control points from collected patient data. The trained model can be used to provide a radiation treatment plan, such as in real-time.

System and method for inspecting a rail using machine learning

An aspect includes a vehicle that includes rail inspection sensors configured for capturing transducer data describing the rail, and a processor configured for receiving and processing the transducer data in near-real time to determine whether the captured transducer data identifies a suspected rail flaw. The processing includes inputting the captured transducer data to a machine learning system that has been trained to identify patterns in transducer data that indicate rail flaws. The processing also includes receiving an output from the machine learning system, the output indicating whether the captured transducer data identifies a suspected rail flaw. An alert is transmitted to an operator of the vehicle based at least in part on the output indicating that the captured transducer data identifies a suspected rail flaw. The alert includes a location of the suspected rail flaw and instructs the operator to stop the vehicle and to perform a repair action.

Method for testing air traffic management electronic system, associated electronic device and platform

Method for testing air traffic management electronic system, associated electronic device and platform The invention relates to an air traffic management electronic system (2), including the steps of: - reception by said system (2) of input data representative of the state of air traffic; - establishment by said system of information relative to the air traffic as a function of said input data and delivery of said information to an electronic test device (6) of the system; - determination by said electronic test device (6), as a function of the delivered information, of air traffic control instructions and providing said system with said instructions; - reception and processing of said instructions by said system; according to which said electronic device (6) includes an algorithmic model (63) for automatically determining instructions as a function of information relative to the air traffic, said model having been obtained during a learning phase, carried out by computer, of a deep learning ...

CONTINUOUS BAYESIAN ESTIMATION WITH A NEURAL NETWORK ARCHITECTURE

Biological neural network

DUPLICATE AND SIMILAR BUG REPORT DETECTION AND RETRIEVAL USING NEURAL NETWORKS

A device, including one or more processors that receive information associated with a first bug report and a second bug report for classification as duplicate bug reports or non duplicate bug reports, identify a first set of descriptions, associated with the first bug report, and a second set of descriptions associated with the second bug report, each description, included in the first set of descriptions, and each corresponding description, included in the second set of descriptions, sharing a description type, identify a neural network, of a plurality of different types of neural networks, for encoding the set of descriptions and the second set of descriptions, based on whether the shared description type is an unstructured data type, a short description type, a long description type, or a structured description type, encode the first set of descriptions into a first set of vectors using the neural network, encode the second set of descriptions into a second set of vectors using the neural ...

Improved artificial digital neuron, neural network and network training algorithm

RADAR HRRP TARGET CLASS LABELING METHOD BASED ON CONVOLUTIONAL AUTO-ENCODER

Aiming at the problems of low efficiency and low accuracy of manual labeling for massive HRRP samples, the present disclosure provides a radar HRRP target class labeling method based on convolutional auto-encoder. The method mainly comprises three stages. In the first stage, a convolutional auto-encoder is constructed, and all HRRP samples are used to train a convolutional auto-encoder to convergence. In the second stage, an encoder of the convolutional auto-encoder is used as a feature extractor to construct the convolutional neural network, and the labeled HRRP samples are used to train the convolutional neural network to obtain the initial labeling model. In the third stage, the unlabeled HRRP samples whose one-hot vector meets the labeling conditions are labeled, and are used to update the parameters of the labeling model. The third stage is repeated until the number of labeled HRRP samples is no longer increased, thus obtaining the final labeling model. Compared with traditional labeling ...

NEUROCOMPUTER

Artificial neural device

LEARNING PROCESSING SYSTEM OF NETWORK STRUCTURE DATA PROCESSING UNIT

END-TO-END SPEAKER RECOGNITION USING DEEP NEURAL NETWORK

COMPUTERIZED ASSISTANCE USING ARTIFICIAL INTELLIGENCE KNOWLEDGE BASE

A computerized personal assistant includes a natural language user interface, a natural language processing machine, an identity machine, and a knowledge-base updating machine. The knowledge-base updating machine is configured to update a user-centric artificial intelligence knowledge base associated with the particular user to include a new or updated user-centric fact based on the computer-readable representation of the user input, wherein the knowledge-base updating machine updates the user-centric artificial intelligence knowledge base via an update protocol useable by a plurality of different computer services.

METHOD FOR DISTINGUISHING A REAL THREE-DIMENSIONAL OBJECT FROM A TWO-DIMENSIONAL SPOOF OF THE REAL OBJECT

A method for distinguishing a real three-dimensional object, like a finger of a hand, from a two- dimensional spoof of the real object, the method comprising: obtaining, by an optical sensor of a mobile device, an image, wherein the image contains either the spoof or the real object; providing the image to a neural network; processing the image by the neural network; wherein processing comprises calculating at least one of: a distance map representative of the distance of a plurality of pixels to the optical sensor, the pixels constituting at least a portion of the object within the image; a reflection pattern representative of light reflection associated with a plurality of pixels constituting at least a portion of the object within the image; and wherein processing further comprises comparing at least one of the calculated distance map or the calculated reflection pattern with a learned distance map or a learned reflection pattern, thereby determining, based on an outcome of the comparison ...

METHOD FOR IDENTIFYING AN OBJECT WITHIN AN IMAGE AND MOBILE DEVICE FOR EXECUTING THE METHOD

A method for identifying a user using an image of an object of the user that has a biometric characteristic of the user, like a fingerprint or a set of fingerprints of fingertips, the method comprising: obtaining, by an optical sensor of a mobile device, the image of the object; providing the image to a neural network; processing the image by the neural network, thereby identifying both, the position of the object and the object in the image; extracting, from the identified object, the biometric characteristic; storing the biometric characteristic in a storage device and/or providing at least the biometric characteristic as input to an identification means, comprising processing the input in order to determine whether the biometric characteristic identifies the user.

NEURAL NETWORK PROCESSOR USING COMPRESSION AND DECOMPRESSION OF ACTIVATION DATA TO REDUCE MEMORY BANDWIDTH UTILIZATION

A deep neural network ("DNN") module can compress and decompress neuron-generated activation data to reduce the utilization of memory bus bandwidth. The compression unit can receive an uncompressed chunk of data generated by a neuron in the DNN module. The compression unit generates a mask portion and a data portion of a compressed output chunk. The mask portion encodes the presence and location of the zero and non-zero bytes in the uncompressed chunk of data. The data portion stores truncated non-zero bytes from the uncompressed chunk of data. A decompression unit can receive a compressed chunk of data from memory in the DNN processor or memory of an application host. The decompression unit decompresses the compressed chunk of data using the mask portion and the data portion. This can reduce memory bus utilization, allow a DNN module to complete processing operations more quickly, and reduce power consumption.

SYSTEMS AND METHODS FOR LEGAL CLAUSE MATCHING AND EXPLANATION

A tool configured to cause the system to perform steps of a method is presented. The method includes receiving labeled training data comprising a labeled set of caselaw. The method further includes training a recurrent neural network model using the labeled training data to generate logical rules, wherein the logical rules comprise rules relating legal clauses from the labeled set of caselaw to outcomes from the labeled set of caselaw. The method includes applying the recurrent network model to a corpus of caselaw to generate a first set logical rules. The method includes receiving a first legal document comprising one or more legal clauses and applying the recurrent network model to the first legal document to generate a second set of logical rules. Based on a comparison of the first set of logical rules with the second set of logical rules, determining a relevant case from the corpus of caselaw.

DEEP-LEARNING-BASED SYSTEM AND PROCESS FOR IMAGE RECOGNITION

Disclosed are methods and systems for using artificial intelligence (AI) for image recognition by using predefined coordinates to extract a portion of a received image, the extracted portion comprising a word to be identified having at least a first letter and a second letter; executing an image recognition protocol to identify the first letter; when the server is unable to identify the second letter, the server executes an AI model having a nodal data structure to identify the second letter based upon the identified first letter, the nodal data structure comprising a set of nodes where each node represents a letter, each node connected to at least one other node, wherein connection of a first node to a second node corresponds to a probability that a letter corresponding to the second node is used in a word subsequent to a letter corresponding to the first node.

FINITE RANK DEEP KERNEL LEARNING WITH LINEAR COMPUTATIONAL COMPLEXITY

Certain aspects of the present disclosure provide techniques for performing finite rank deep kernel learning. In one example, a method for performing finite rank deep kernel learning includes receiving a training dataset; forming a set of embeddings by subjecting the training dataset to a deep neural network; forming, from the set of embeddings, a plurality of dot kernels; linearly combining the plurality of dot kernels to form a composite kernel for a Gaussian process; receiving live data from an application; and predicting a plurality of values and a plurality of uncertainties associated with the plurality of values simultaneously using the composite kernel.

NEURAL NETWORK PROCESSOR

A circuit for performing computations for a neural network comprising multiple neural network (NN) layers. The circuit includes a processing device that provides programming data for performing the computations and a core in data communication with the processing device to receive the programming data. The core includes activation memory that stores inputs for a layer and parameter memory that stores parameters for a first NN layer. The core also includes a rotation unit that rotates accessing the inputs from the activation memory based on the programming data and a computation unit that receives a respective input and a parameter for the first NN layer and generates an output of the first NN layer using the input and the parameter. The core also includes a crossbar unit that causes the output to be stored, in the activation memory, in accordance with a bank assignment pattern.

DETECTING INFECTION OF PLANT DISEASES WITH IMPROVED MACHINE LEARNING

A system and processing methods for refining a convolutional neural network (CNN) to capture characterizing features of different classes are disclosed. In some embodiments, the system is programmed to start with the filters in one of the last few convolutional layers of the initial CNN, which often correspond to more class-specific features, rank them to hone in on more relevant filters, and update the initial CNN by turning off the less relevant filters in that one convolutional layer. The result is often a more generalized CNN that is rid of certain filters that do not help characterize the classes.

MODEL-FREE CONTROL OF DYNAMICAL SYSTEMS WITH DEEP RESERVOIR COMPUTING

A technique is provided for control of a nonlinear dynamical system to an arbitrary trajectory. The technique does not require any knowledge of the dynamical system, and thus is completely model-free. When applied to a chaotic system, it is capable of stabilizing unstable periodic orbits (UPOs) and unstable steady states (USSs), controlling orbits that require non-vanishing control signal, synchronization to other chaotic systems, and so on. It is based on a type of recurrent neural network (RNN) known as a reservoir computer (RC), which, as shown, is capable of directly learning how to control an unknown system. Precise control to a desired trajectory is obtained by iteratively adding layers to the controller, forming a deep recurrent neural network.

MEDICAL ROBOT COMPRISING AUTOMATIC POSITIONING MEANS

L'invention concerne un robot médical (10) comportant une base mobile (13) motorisée, des capteurs (17) de repérage dans l'espace solidaires de la base mobile, et une unité de contrôle (16) mémorisant un plan d'intervention comportant au moins une action à réaliser sur une anatomie d'intérêt d'un patient (30). L'unité de contrôle est configurée pour : - détecter, à partir d'informations provenant des capteurs (17) de repérage dans l'espace, une position de l'anatomie d'intérêt du patient par rapport au robot médical, - identifier, à partir de la position de l'anatomie d'intérêt du patient et du plan d'intervention, au moins une position favorable de la base mobile du robot médical pour laquelle le robot médical est capable d'effectuer la ou les actions du plan d'intervention, - déplacer la base mobile du robot médical à une position optimale sélectionnée parmi la ou les positions favorables identifiées.

STRUCTURE DEFECT DETECTION USING MACHINE LEARNING ALGORITHMS

Structure defect detection is performed using computer-implemented arrangements employing machine learning algorithms in the form of neural networks. In one arrangement, a convolutional neural network is trained using a database of images formed to optimize accuracy of the convolutional neural network to detect, for example, a crack in a concrete surface. A two-stage scanning process each performing a plurality of scans of a test image is incorporated in the foregoing arrangement of convolutional neural network, with the two-stages forming overlapping capture areas to reduce likelihood of a crack lying on a boundary of the individual scans going undetected. Also, region-based convolutional neural networks are trained to detect various types of defects.

Methods and systems for memristor-based neuron circuits

Certain embodiments of the present disclosure support techniques for designing neuron circuits based on memristors. Bulky capacitors as electrical current integrators can be eliminated and nanometer scale memristors can be utilized instead. Using the nanometer feature-sized memristors, the neuron hardware area can be substantially reduced.

Systems and methods for processing data flows

A flow processing facility, which uses a set of artificial neurons for pattern recognition, such as a self-organizing map, in order to provide security and protection to a computer or computer system supports unified threat management based at least in part on patterns relevant to a variety of types of threats that relate to computer systems, including computer networks. Flow processing for switching, security, and other network applications, including a facility that processes a data flow to address patterns relevant to a variety of conditions are directed at internal network security, virtualization, and web connection security. A flow processing facility for inspecting payloads of network traffic packets detects security threats and intrusions across accessible layers of the IP-stack by applying content matching and behavioral anomaly detection techniques based on regular expression matching and self-organizing maps. Exposing threats and intrusions within packet payload at or near real-time rates enhances network security from both external and internal sources while ensuring security policy is rigorously applied to data and system resources. Intrusion Detection and Protection (IDP) is provided by a flow processing facility that processes a data flow to address patterns relevant to a variety of types of network and data integrity threats that relate to computer systems, including computer networks.

Systems, methods, and apparatus for otoacoustic protection of autonomic systems

Systems, methods and apparatus are provided through which in some embodiments an autonomic unit transmits an otoacoustic signal to counteract a potentially harmful incoming signal.

Deep belief network for large vocabulary continuous speech recognition

A method is disclosed herein that includes an act of causing a processor to receive a sample, wherein the sample is one of spoken utterance, an online handwriting sample, or a moving image sample. The method also comprises the act of causing the processor to decode the sample based at least in part upon an output of a combination of a deep structure and a context-dependent Hidden Markov Model (HMM), wherein the deep structure is configured to output a posterior probability of a context-dependent unit. The deep structure is a Deep Belief Network consisting of many layers of nonlinear units with connecting weights between layers trained by a pretraining step followed by a fine-tuning step.

Bioinspired System for Image Processing

A method for digital image processing is bioinspired and includes an architecture that emulates the functions of photoreceptors, horizontal cells, bipolar cells and ganglion cells of a primate retina based on an image as input. The method detects edges and properties of the surfaces present in the digital image. The output is a data set that includes photoreceptor emulators that emulate photoreceptor cells and connected to the data input. Each emulator includes a cellular base structure with a modulated data input, a calculation center to process the modulated data and an output of the data processed by the calculation center, and the emulators forming a virtual retina in which each emulator is parameterized.

Accurate and Fast Neural network Training for Library-Based Critical Dimension (CD) Metrology

Approaches for accurate neural network training for library-based critical dimension (CD) metrology are described. Approaches for fast neural network training for library-based CD metrology are also described.

Electronic brain model with neuron tables

A method of emulating the human brain with its thought and rationalization processes is presented here, as well as a method of storing human-like thought. The invention provides for inclusion of psychological profiles, experience and societal position in an electronic emulation of the human brain. This permits a realistic human-like response by that emulation to the people and the interactive environment around it.

Reconfigurable and customizable general-purpose circuits for neural networks

A reconfigurable neural network circuit is provided. The reconfigurable neural network circuit comprises an electronic synapse array including multiple synapses interconnecting a plurality of digital electronic neurons. Each neuron comprises an integrator that integrates input spikes and generates a signal when the integrated inputs exceed a threshold. The circuit further comprises a control module for reconfiguring the synapse array. The control module comprises a global final state machine that controls timing for operation of the circuit, and a priority encoder that allows spiking neurons to sequentially access the synapse array.

Method and apparatus for event detection permitting per event adjustment of false alarm rate

Method and apparatus for object or event of interest detection which minimizes the level of false alarms and maximizes the level of detections as defined on a per event or object basis by the analyst. The invention allows for the minimization of false alarms for objects or events of interest which have a close resemblance to all other objects or events mapped to the same multidimensional feature space, and allows for the per event or per object adjustment on false alarms for objects or events of higher interest.

Time encoding using integrate and fire sampler

Systems and methods of time encoding using an integrate and fire (IF) sampler are disclosed. In an example, a method includes receiving input signals for separate classes. The method also includes generating a pulse train based on the input signals. The method also includes binning the pulse train to generate a feature vector.

Synapse for function cell of spike timing dependent plasticity (stdp), function cell of stdp, and neuromorphic circuit using function cell of stdp

A synapse for a spike timing dependent (STDP) function cell includes a memory device having a variable resistance, such as a memristor, and a transistor connected to the memory device. A channel of the memory device is connected in series with a channel of the transistor.

Neural network system and method for controlling output based on user feedback

For various information sources, information output based on user feedback about information from the sources is controlled. A neural network module selects object(s) to receive information from the information sources based on inputs and weight values during that epoch. A server, associated with the neural network module, provides the object(s) to recipients. The object(s) may comprise electronic mail messages, chat participants viewers, or slots within a link directory page. The recipients provide feedback about the information during an epoch. At the conclusion of an epoch, the neural network takes the feedback provided by the recipients and generates a rating value for the object(s). Based on the rating value and the selections made, the neural network re-determines the weight values within the network. The neural network then selects the object(s) to receive information during a subsequent epoch using the re-determined weight values and the inputs for that subsequent epoch.

Method and apparatus of robust neural temporal coding, learning and cell recruitments for memory using oscillation

Certain aspects of the present disclosure support a technique for robust neural temporal coding, learning and cell recruitments for memory using oscillations. Methods are proposed for distinguishing temporal patterns and, in contrast to other “temporal pattern” methods, not merely coincidence of inputs or order of inputs. Moreover, the present disclosure propose practical methods that are biologically-inspired/consistent but reduced in complexity and capable of coding, decoding, recognizing, and learning temporal spike signal patterns. In this disclosure, extensions are proposed to a scalable temporal neural model for robustness, confidence or integrity coding, and recruitment of cells for efficient temporal pattern memory.

Unsupervised, supervised, and reinforced learning via spiking computation

The present invention relates to unsupervised, supervised and reinforced learning via spiking computation. The neural network comprises a plurality of neural modules. Each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. An interconnection network comprising a plurality of edges interconnects the plurality of neural modules. Each edge interconnects a first neural module to a second neural module, and each edge comprises a weighted synaptic connection between every neuron in the first neural module and a corresponding neuron in the second neural module.

Discriminative pretraining of deep neural networks

Discriminative pretraining technique embodiments are presented that pretrain the hidden layers of a Deep Neural Network (DNN). In general, a one-hidden-layer neural network is trained first using labels discriminatively with error back-propagation (BP). Then, after discarding an output layer in the previous one-hidden-layer neural network, another randomly initialized hidden layer is added on top of the previously trained hidden layer along with a new output layer that represents the targets for classification or recognition. The resulting multiple-hidden-layer DNN is then discriminatively trained using the same strategy, and so on until the desired number of hidden layers is reached. This produces a pretrained DNN. The discriminative pretraining technique embodiments have the advantage of bringing the DNN layer weights close to a good local optimum, while still leaving them in a range with a high gradient so that they can be fine-tuned effectively.

ELECTRONIC CIRCUIT WITH NEUROMORPHIC ARCHITECTURE

Neuromorphic circuits are multi-cell networks configured to imitate the behavior of biological neural networks. A neuromorphic circuit is provided which comprises a network of neurons each identified by a neuron address in the network, each neuron being able to receive and process at least one input signal and then later emit on an output of the neuron a signal representing an event which occurs inside the neuron, and a programmable memory composed of elementary memories each associated with a respective neuron. The elementary memory, which is a memory of post-synaptic addresses and weights, comprises an activation input linked by a conductor to the output of the associated neuron to directly receive an event signal emitted by this neuron without passing through an address encoder or decoder. The post-synaptic addresses extracted from an elementary memory activated by a neuron are applied, with associated synaptic weights, as inputs to the neural network. 2. The neuromorphic circuit as claimed in claim 1 , wherein the logic circuit of the elementary memory comprises a conflicts management circuit for preventing the writing of data to the bus while a datum arising from another elementary memory is being emitted on the bus.3. The neuromorphic circuit as claimed in claim 1 , wherein the elementary memory is juxtaposed with the associated neuron in the neural network.4. The neuromorphic circuit as claimed in claim 1 , wherein the elementary memory is situated above the associated neuron.5. The neuromorphic circuit as claimed in claim 4 , wherein the neural network is situated on a first integrated circuit and the memory on a second integrated circuit placed above the first claim 4 , the elementary memory being linked to the associated neuron by a respective contact between the two chips.6. The neuromorphic circuit as claimed in claim 1 , wherein the elementary memory comprises several memory registers claim 1 , each register containing an address and an associated ...

NEURAL NETWORK DESIGNING METHOD AND DIGITAL-TO-ANALOG FITTING METHOD

A neural network designing method forms a RNN (Recurrent Neural Network) circuit to include a plurality of oscillating RNN circuits configured to output natural oscillations, and an adding circuit configured to obtain a sum of outputs of the plurality of oscillating RNN circuits, and inputs discrete data to the plurality of oscillating RNN circuits in order to compute a fitting curve with respect to the discrete data output from the adding circuit. 1. A neural network designing method comprising:a forming procedure causing a computer to form a RNN (Recurrent Neural Network) circuit to include a plurality of oscillating RNN circuits configured to output natural oscillations, and an adding circuit configured to obtain a sum of outputs of the plurality of oscillating RNN circuits; anda computing procedure causing the computer to input discrete data to the plurality of oscillating RNN circuits, in order to compute a fitting curve with respect to the discrete data output from the adding circuit.2. The neural network designing method as claimed in claim 1 , wherein claim 1 , when an actual data length of the discrete data is an odd number 2n−1 claim 1 , where n is a natural number greater than or equal to 2 claim 1 , the forming procedure sets a data length of the discrete data to an even number 2n greater than the actual data length.3. The neural network designing method as claimed in claim 1 , wherein the discrete data are acceleration data claim 1 , and the fitting curve is an acceleration fitting curve or a locus curve.6. A non-transitory computer-readable storage medium having stored therein a program for causing a computer to execute a neural network designing process comprising:forming a RNN (Recurrent Neural Network) circuit to include a plurality of oscillating RNN circuits configured to output natural oscillations, and an adding circuit configured to obtain a sum of outputs of the plurality of oscillating RNN circuits; andcomputing procedure causing the computer ...

Retina prosthesis

This disclosure provides a retinal prosthetic method and device that mimics the responses of the retina to a broad range of stimuli, including natural stimuli. Ganglion cell firing patterns are generated in response to a stimulus using a set of encoders, interfaces, and transducers, where each transducer targets a single cell or a small number of cells. The conversion occurs on the same time scale as that carried out by the normal retina. In addition, aspects of the invention may be used with robotic or other mechanical devices, where processing of visual information is required. The encoders may be adjusted over time with aging or the progression of a disease.

SYSTEMS, METHODS AND COMPUTER-READABLE MEDIA FOR GENERATING JUDICIAL PREDICTION INFORMATION

Systems, methods and computer-readable storage media for generating judicial prediction information are described. A judicial information prediction system may be configured to receive an analysis request comprising judicial request elements and to access at least one judicial information source associated with the analysis request. The judicial request elements may include items of interest associated with the prediction of a legal action. For example, the judicial request elements may include a court, a judge, an area of the law, and a legal action, such as a motion to dismiss. The judicial information prediction system may operate to analyze the at least one judicial information source based on the judicial request elements to generate judicial prediction information. For instance, the judicial prediction information may indicate the likelihood of success of a legal event based on the circumstances specified in the analysis request. 1. A judicial prediction information system comprising:a processor; anda non-transitory, computer-readable storage medium in operable communication with the processor, wherein the computer-readable storage medium contains one or more programming instructions that, when executed, cause the processor to:receive an analysis request comprising judicial request elements;access at least one judicial information source associated with the analysis request; andanalyze the at least one judicial information source based on the judicial request elements to generate judicial prediction information.2. The system of claim 1 , wherein the judicial request elements comprise at least one of the following: a judicial entity claim 1 , a judicial actor claim 1 , a case type claim 1 , a case issue claim 1 , a procedural posture claim 1 , at least one keyword claim 1 , and at least one factor.3. The system of claim 1 , wherein the judicial request elements comprise at least one judicial entity claim 1 , at least one judicial actor and at least one case ...

LEARNING METHOD OF NEURAL NETWORK CIRCUIT

A neuron circuit in a neural network circuit element includes a waveform generating circuit for generating a bipolar sawtooth pulse voltage, and a first input signal has a bipolar sawtooth pulse waveform. For a period during which the first input signal is permitted to be input to a first electrode of a variable resistance element, the bipolar sawtooth pulse voltage generated within the neural network circuit element including the variable resistance element which is applied with the first input signal from another neural network circuit element is input to a control electrode of the variable resistance element. The resistance value of the variable resistance element changes due to an electric potential difference between the first electrode and the control electrode, the electric potential difference being generated depending on an input timing difference between a voltage applied to the first electrode and the voltage applied to the control electrode. 1. A learning method of a neural network circuit including a plurality of neural network circuit elements which are interconnected ,wherein each of the plurality of neural network circuit elements includes:at least one synapse circuit which receives as an input a signal (first input signal) output from another neural network circuit element; andone neuron circuit which receives as an input a signal output from the at least one synapse circuit,wherein the synapse circuit includes a variable resistance element which includes a first electrode formed on and above a semiconductive layer; a second electrode formed on and above the semiconductive layer; and a control electrode formed on a main surface of the semiconductive layer via a ferroelectric layer and changes a resistance value between the first electrode and the second electrode in response to an electric potential difference between the first electrode and the control electrode;wherein the synapse circuit is configured to perform switching between a state in which ...

Neural network-based turbine monitoring system

A neural network-based system for monitoring a turbine compressor. In various embodiments, the neural network-based system includes: at least one computing device configured to monitor a turbine compressor by performing actions including: comparing a monitoring output from a first artificial neural network (ANN) about the turbine compressor to a monitoring output from a second, distinct ANN about the turbine compressor; and predicting a probability of a malfunction in the turbine compressor based upon the comparison of the monitoring outputs from the first ANN and the second, distinct ANN.

ANALOG PROGRAMMABLE SPARSE APPROXIMATION SYSTEM

A system and device for solving sparse algorithms using hardware solutions is described. The hardware solution can comprise one or more analog devices for providing fast, energy efficient solutions to small, medium, and large sparse approximation problems. The system can comprise sub-threshold current mode circuits on a Field Programmable Analog Array (FPAA) or on a custom analog chip. The system can comprise a plurality of floating gates for solving linear portions of a sparse signal. The system can also comprise one or more analog devices for solving non-linear portions of sparse signal. 1. A method comprising:applying each of a plurality of input signals to each of a plurality of feedforward excitation signals to generate a plurality of first output signals;applying each of a plurality of second output signals to each of a plurality of lateral inhibition signals to generate a plurality of recurrent feedback signals;subtracting each the plurality of recurrent feedback signals from each of the plurality of first output signals to generate a plurality of intermediate signals; andapplying each of the plurality of intermediate signals to a non-linear computation to generate the plurality of second output signals.2. The method of claim 1 , further comprising:converting a first sparse vector of a plurality of sparse vectors to a plurality of input signals.3. The method of claim 1 , wherein the plurality of feedforward excitation signals are applied by a first plurality of transistors that comprise a first analog vector matrix multiplier (VMM).4. The method of claim 1 , wherein the plurality of lateral inhibition signals are applied by a second plurality of transistors that comprise a second analog vector matrix multiplier (VMM).5. The method of claim 1 , wherein the subtraction step is performed by a plurality of current mirrors.6. The method of claim 1 , wherein each step is performed in parallel in continuous time for each input signal of the plurality of input ...

Cascading learning system as semantic search

A cascading learning system as a semantic search is described. The cascading learning system has a request analyzer, a request dispatcher and classifier, a search module, a terminology manager, and a cluster manager. The request analyzer receives a request for search terms from a client application and determines term context in the request to normalize request data from the term context. The normalized request data are classified and dispatched to a corresponding domain-specific module. Each domain-specific module of a search module generates a prediction with a trained probability of an expected output. The terminology manager receives normalized request data from the request dispatcher and classifier, and manages terminology stored in a contextual network. The cluster manager controls data flow between the request dispatcher and classifier, the search module container, the terminology manager, and a business data source system.

NEUROMORPHIC SIGNAL PROCESSING DEVICE AND METHOD FOR LOCATING SOUND SOURCE USING A PLURALITY OF NEURON CIRCUITS

Provided is a neuromorphic signal processing device for locating a sound source using a plurality of neuron circuits, the neuromorphic signal processing device including a detector configured to output a detected spiking signal using a detection neuron circuit corresponding to a predetermined time difference, in response to a first signal and a second signal containing an identical input spiking signal with respect to the predetermined time difference, for each of a plurality of predetermined frequency bands, a multiplexor configured to output a multiplexed spiking signal corresponding to the predetermined time difference based on a plurality of the detected spiking signals output from a plurality of neuron circuits corresponding to the plurality of frequency bands, and an integrator configured to output an integrated spiking signal corresponding to the predetermined time difference, based on a plurality of the multiplexed spiking signals corresponding to a plurality of predetermined time differences. 1. A neuromorphic signal processing device for locating a sound source using a plurality of neuron circuits , the device comprising:a detector configured to output a detected spiking signal using a detection neuron circuit corresponding to a predetermined time difference, in response to a first signal and a second signal containing an identical input spiking signal with respect to the predetermined time difference, for each of a plurality of predetermined frequency bands;a multiplexor configured to output a multiplexed spiking signal corresponding to the predetermined time difference based on a plurality of the detected spiking signals output from a plurality of neuron circuits corresponding to the plurality of frequency bands; andan integrator configured to output an integrated spiking signal corresponding to the predetermined time difference, based on a plurality of the multiplexed spiking signals corresponding to a plurality of predetermined time differences,wherein ...

Learning spike timing precision

Certain aspects of the present disclosure provide methods and apparatus for learning or determining delays between neuron models so that the uncertainty in input spike timing is accounted for in the margin of time between a delayed pre-synaptic input spike and a post-synaptic spike. In this manner, a neural network can correctly match patterns (even in the presence of significant jitter) and correctly distinguish between different noisy patterns. One example method generally includes determining an uncertainty associated with a first pre-synaptic spike time of a first neuron model for a pattern to be learned; and determining a delay based on the uncertainty, such that the delay added to a second pre-synaptic spike time of the first neuron model results in a causal margin of time between the delayed second pre-synaptic spike time and a post-synaptic spike time of a second neuron model.

UNIVERSAL, ONLINE LEARNING IN MULTI-MODAL PERCEPTION-ACTION SEMILATTICES

In one embodiment, the present invention provides a method for interconnecting neurons in a neural network. At least one node among a first set of nodes is interconnected to at least one node among a second set of nodes, and nodes of the first and second set are arranged in a lattice. At least one node of the first set represents a sensory-motor modality of the neural network. At least one node of the second set is a union of at least two nodes of the first set. Each node in the lattice has an acyclic digraph comprising multiple vertices and directed edges. Each vertex represents a neuron population. Each directed edge comprises multiple synaptic connections. Vertices in different acyclic digraphs are interconnected using an acyclic bottom-up digraph. The bottom-up digraph has a corresponding acyclic top-down digraph. Vertices in the bottom-up digraph are interconnected to vertices in the top-down digraph. 126-. (canceled)27. A signaling interconnect for neural nodes , comprising:an interconnection lattice that interconnects a plurality of first nodes in a first set of nodes with a plurality of second nodes in a second set of nodes, wherein the connected nodes are arranged in a lattice such that a connected node in the second set is a union of at least two nodes in the first set, and said connected node exchanges signals with said at least two nodes via the interconnection lattice.28. The signaling interconnect of claim 27 , wherein:a node comprises one or more neural vertices interconnected via multiple directed edges arranged in an acyclic digraph, wherein each neural vertex comprises one or more neurons, and each edge comprises a signaling pathway in the interconnection lattice.29. The signaling interconnect of claim 28 , wherein:the interconnection lattice interconnects said nodes via bottom-up signaling pathways arranged in an acyclic bottom-up digraph in the interconnection lattice, each bottom-up signaling pathway including one or more vertices and directed ...

Determination of subsurface properties of a well

Embodiments of the present invention provide techniques for using data from a select set of wells to develop correlations between surface-measured properties and properties typically determined from subsurface measurements (e.g., from logging tool responses, core analysis, or other subsurface measurements). When new wells are drilled, the surface data acquired while drilling may be used as an input to these correlations in order to predict properties associated with subsurface measurements.

Apparatus and methods for reinforcement learning in large populations of artificial spiking neurons

Neural network apparatus and methods for implementing reinforcement learning. In one implementation, the neural network is a spiking neural network, and the apparatus and methods may be used for example to enable an adaptive signal processing system to effect network adaptation by optimized credit assignment. In certain implementations, the credit assignment may be based on a comparison between network output and individual unit contribution. The unit contribution may be determined for example using eligibility traces that may comprise pre-synaptic and/or post-synaptic activity. In certain implementations, the unit credit may be determined using correlation between rate of change of network output and eligibility trace of the unit.

Apparatus and methods for efficient updates in spiking neuron network

Efficient updates of connections in artificial neuron networks may be implemented. A framework may be used to describe the connections using a linear synaptic dynamic process, characterized by stable equilibrium. The state of neurons and synapses within the network may be updated, based on inputs and outputs to/from neurons. In some implementations, the updates may be implemented at regular time intervals. In one or more implementations, the updates may be implemented on-demand, based on the network activity (e.g., neuron output and/or input) so as to further reduce computational load associated with the synaptic updates. The connection updates may be decomposed into multiple event-dependent connection change components that may be used to describe connection plasticity change due to neuron input. Using event-dependent connection change components, connection updates may be executed on per neuron basis, as opposed to per-connection basis.

Multi-compartment neurons with neural cores

Embodiments of the invention provide a neural core circuit comprising a synaptic interconnect network including plural electronic synapses for interconnecting one or more source electronic neurons with one or more target electronic neurons. The interconnect network further includes multiple axon paths and multiple dendrite paths. Each synapse is at a cross-point junction of the interconnect network between a dendrite path and an axon path. The core circuit further comprises a routing module maintaining routing information. The routing module routes output from a source electronic neuron to one or more selected axon paths. Each synapse provides a configurable level of signal conduction from an axon path of a source electronic neuron to a dendrite path of a target electronic neuron.

SYNAPTIC, DENDRITIC, SOMATIC, AND AXONAL PLASTICITY IN A NETWORK OF NEURAL CORES USING A PLASTIC MULTI-STAGE CROSSBAR SWITCHING

Embodiments of the invention provide a neural network comprising multiple functional neural core circuits, and a dynamically reconfigurable switch interconnect between the functional neural core circuits. The interconnect comprises multiple connectivity neural core circuits. Each functional neural core circuit comprises a first and a second core module. Each core module comprises a plurality of electronic neurons, a plurality of incoming electronic axons, and multiple electronic synapses interconnecting the incoming axons to the neurons. Each neuron has a corresponding outgoing electronic axon. In one embodiment, zero or more sets of connectivity neural core circuits interconnect outgoing axons in a functional neural core circuit to incoming axons in the same functional neural core circuit. In another embodiment, zero or more sets of connectivity neural core circuits interconnect outgoing and incoming axons in a functional neural core circuit to incoming and outgoing axons in a different functional neural core circuit, respectively. 1. A neural network , comprising:multiple functional neural core circuits; anda dynamically reconfigurable switch interconnect between said multiple functional neural core circuits, wherein the switch interconnect comprises multiple connectivity neural core circuits.2. The neural network of claim 1 , wherein: a plurality of electronic neurons, wherein each neuron has a corresponding outgoing electronic axon;', 'a plurality of incoming electronic axons; and', 'a plurality of intra-core electronic synapses interconnecting the incoming axons to the neurons, wherein each synapse interconnects an incoming axon to a neuron such that each neuron receives axonal firing events from interconnected incoming axons and generates a neuronal firing event according to a neuronal activation function;', 'wherein the first neural core module and the second neural core module are logically overlayed on one another such that neurons in the first neural core ...

METHOD AND SYSTEM FOR WASTEWATER TREATMENT BASED ON DISSOLVED OXYGEN CONTROL BY FUZZY NEURAL NETWORK

A method and system for wastewater treatment based on dissolved oxygen control by a fuzzy neural network, the method for wastewater treatment comprising the following steps: (1) measuring art inlet water flow rate, an ORP value in an anaerobic tank, a DO value in an aerobic tank, an inlet water COD value, and an actual outlet water COD value; (2) collecting the measured sample data and sending them via a computer to a COD fuzzy neural network predictive model, so as to establish an outlet water COD predicted value, (3) comparing the outlet COD predicted value with the outlet water COD set value, so as to obtain an error and an error change rate, and using them as two input variables to adjust a suitable dissolved oxygen concentration. Accordingly, the on-line prediction and real-time control of dissolved oxygen wastewater treatment are achieved. The accurate control of dissolved oxygen concentration by the present method for wastewater treatment can achieve a saving in energy consumption while ensuring stable running of the sewage treatment system, and the outlet water quality meets the national emission standards. 1. A method of wastewater treatment based on dissolved oxygen control by fuzzy neural network comprising:(1) Measuring an inflow flowrate, an ORP value in an anaerobic tank corresponding to the real-time aeration quantity, a DO value in an aerobic tank corresponding to the real-time aeration quantity, an influent COD value, and an actual effluent COD value in the A/O wastewater treatment process;(2) Collecting the measured sampling data, sending them via a computer to a COD fuzzy neural network predictive model, and computing as physical quantities, so as to establish an effluent COD predicted value;(3) Comparing the effluent water COD predicted value with the effluent COD set value, an error and an error change rate of the effluent COD value are obtained. And the error and the error change rate of the effluent COD value are used as two input variables of ...

Wellbore completion and hydraulic fracturing optimization methods and associated systems

Methods and systems for optimizing wellbore completion and, in particular, methods and systems for optimizing hydraulic fracturing parameters are disclosed. In some embodiments, a method of optimizing wellbore completion includes gathering wellbore data, screening and processing the gathered wellbore data, utilizing the screened and processed wellbore data to define an optimized model, and utilizing the optimized model to evaluate combinations of available wellbore completion parameters. In some instances, the optimized model is defined using artificial neural networks, genetic algorithms, and/or boosted regression trees. Further, in some embodiments the combinations of available wellbore completion parameters include hydraulic fracturing parameters, such as number of fractures, fracturing fluid type, proppant type, fracturing volume, and/or other parameters.

COMPUTER-IMPLEMENTED SIMULATED INTELLIGENCE CAPABILITIES BY NEUROANATOMICALLY-BASED SYSTEM ARCHITECTURE

Computer-implemented systems for simulated intelligence information processing comprising: a digital processing device comprising an operating system configured to perform executable instructions and a memory; a computer program including instructions executable by the digital processing device to create a hierarchical software architecture comprising: a software module for providing a functional interpretation of the prosencephalon, or parts thereof; a software module for providing a functional interpretation of the mesencephalon, or parts thereof; and a software module for providing a functional interpretation of the rhombencephalon, or parts thereof; wherein the software architecture simulates vertebrate, mammalian, primate, or human neuroanatomy. In some embodiments, the systems create simulated intelligence. 1. A computer-implemented system for simulated intelligence information processing comprising:a. a digital processing device comprising an operating system configured to perform executable instructions and a memory; i. a module for providing a functional interpretation of the prosencephalon;', 'ii. a module for providing a functional interpretation of the mesencephalon; and', 'iii. a module for providing a functional interpretation of the rhombencephalon;, 'b. a computer program including instructions executable by the digital processing device to create a hierarchical software architecture for creation of applications that simulate a brain, the architecture comprisingwherein said hierarchical software architecture simulates the cognitive information processing of vertebrate, mammalian, primate, or human neuroanatomy.2. The system of claim 1 , wherein the module for providing a functional interpretation of the prosencephalon comprises functional interpretations of the telencephalon and diencephalon.3. The system of claim 1 , wherein the module for providing a functional interpretation of the mesencephalon comprises functional interpretations of and inferior ...

HYBRID INTERCONNECT STRATEGY FOR LARGE-SCALE NEURAL NETWORK SYSTEMS

A plurality of chips arranged in a certain layout so as to face free space, and one or more optical elements are included. In the case where signal traffic for electrical communication with a given chip exceeds or is expected to exceed a certain threshold, a plurality of chips involved in communication routing of the excess signal traffic are identified, part of related signal traffic that crosses the plurality of identified chips is converted from an electric signal into an optical signal to re-route the excess signal traffic, and paths of the related signal traffic are dynamically adapted from fixed wired paths between the plurality of chips to optical communication paths formed in the free space. 1. A network system , comprising:a plurality of chips arranged in a certain layout to face free space, a plurality of certain chips among the plurality of chips being configured to be able to electrically communicate with each other via fixed wired paths; andone or more optical elements configured to convert an electric signal of a given chip among the plurality of chips into an optical signal and configured to enable optical communication to another chip via optical communication paths selected in the free space, direct communication from the given chip to the other chip not being electrically established via the fixed wired path;wherein in a case where signal traffic for electrical communication with a given chip exceeds or is expected to exceed a certain threshold, a plurality of chips involved in communication routing of the excess signal traffic are identified, part of related signal traffic that crosses the plurality of identified chips is converted from an electric signal into an optical signal, and the paths of the related signal traffic are dynamically and reconfigurably adapted from the fixed wired paths between the plurality of chips to optical communication paths formed by the one or more optical elements in the free space in order to re-route the excess ...

Neural network learning and collaboration apparatus and methods

Apparatus and methods for learning and training in neural network-based devices. In one implementation, the devices each comprise multiple spiking neurons, configured to process sensory input. In one approach, alternate heterosynaptic plasticity mechanisms are used to enhance learning and field diversity within the devices. The selection of alternate plasticity rules is based on recent post-synaptic activity of neighboring neurons. Apparatus and methods for simplifying training of the devices are also disclosed, including a computer-based application. A data representation of the neural network may be imaged and transferred to another computational environment, effectively copying the brain. Techniques and architectures for achieve this training, storing, and distributing these data representations are also disclosed.

TECHNIQUE FOR EFFICIENT RETRIEVAL OF PERSONALITY DATA

A technique for enabling efficient retrieval of a digital representation of personality data of a user () by a client device () from a server () is disclosed, wherein the digital representation of the personality data is processed at the client device () to provide a user-adapted service to the user (). A method implementation of the technique is performed by the server () and comprises storing a neural network being trained to compute personality data of a user based on input obtained from the user (), receiving, from the client device (), a request for a digital representation of personality data for a user (), and sending, to the client device (), the requested digital representation of the personality data of the user (), wherein the personality data of the user is computed using the neural network based on input obtained from the user (). 1. A method including a retrieval of a digital representation of personality data of a user by a client device from a server , the method being performed by the server and comprising:storing a neural network trained to compute personality data of a user based on input obtained from the user;receiving, from the client device, a request for a digital representation of personality data for a user; andsending, to the client device, the requested digital representation of the personality data of the user, the digital representation of the personality data of the user being processed at the client device to provide a user-adapted service to the user, wherein the personality data of the user is computed using the neural network based on input obtained from the user, and wherein the method further comprises:receiving feedback characterizing the user;updating the neural network based on the feedback; andsending, to the client device, a digital representation of updated personality data of the user, wherein the updated personality data of the user is computed using the updated neural network.2. The method of claim 1 , wherein the ...

MAPPING EFFICIENCY BY SUGGESTING MAP POINT'S LOCATION

A method and apparatus of mapping efficiency by suggesting map points location includes receiving data at a machine, the data including a plurality of signals received during the performance of a triangulation to locate a focal tachycardia, generating, by the machine, a prediction model as to the location of the focal tachycardia, and modifying, by the machine, the prediction model based upon additional data received by the machine. 1. A method of mapping efficiency by suggesting map points location , comprising:receiving data at a machine, the data including a plurality of signals received during the performance of a triangulation to locate a focal tachycardia;generating, by the machine, a prediction model as to the location of the focal tachycardia; andmodifying, by the machine, the prediction model based upon additional data received by the machine.2. The method of wherein local activation points (LATs) are used to estimate the focal point in a triangle.3. The method of wherein the data is anatomy (FAM/CT) data and LAT points acquired by a Carto machine.4. The method of wherein the data includes an ablation location of a focal/termination indication by a physician or a coherent map.5. The method of wherein the data is described at each stage by a single LAT point (position & activation).6. The method of wherein the location of a next best point to sample over an anatomy is represented as coordinates in space.7. The method of claim 1 , further comprising learning and using similarities in patients to further provide a more accurate prediction for a physician.8. The method of wherein the data includes information based on a specific disease.9. The method of claim 1 , wherein the data in an input space is divided into grid sampled voxels with each voxel including the electrical activation signals measured inside the voxel.10. A system for focal point location claim 1 , the system comprising:a plurality of inputs;a first converter that converts at least a first ...

SYSTEMS, METHODS, AND MEDIA FOR AUTOMATICALLY DIAGNOSING INTRADUCTAL PAPILLARY MUCINOUS NEOSPLASMS USING MULTI-MODAL MAGNETIC RESONANCE IMAGING DATA

In accordance with some embodiments, systems, methods, and media for automatically diagnosing IPMNs using multi-modal MRI data are provided. In some embodiments, a system comprises: an MRI scanner; and a processor programmed to: prompt a user to select a slice of T1 and T2 MRI data including the subject's pancreas; generate minimum and maximum intensity projections based consecutive slices of the T1 and T2 MRI data; provide the projections to an image recognition CNN, and receive feature vectors for each from a fully connected layer; perform a canonical correlation analysis to determine correlations between the feature vectors; and provide a resultant vector to an SVM that determines whether the subject's pancreas includes IPMNs based on a vector. 1. A system for automatically detecting the presence of intraductal papillary mucinous neoplasms (IPMNs) in a subject's pancreas , the system comprising: [{'sub': 1', 'u', 'N', {'sub2': '1'}], 'receive T1 MRI data generated by an MRI machine, the T1 MRI data comprising a plurality of slices of T1 MRI data [I, . . . , I, . . . , I];'}, {'sub': 1', 'v', 'N', {'sub2': '2'}], 'receive T2 MRI data generated by the MRI machine, the T2 MRI data comprising a plurality of slices of T2 MRI data [J, . . . , J, . . . , J];'}, 'provide data representing k slices of the T1 MRI data to a trained image classification convolutional neural network (CNN);', 'provide data representing k slices of the T2 MRI data to the trained image classification CNN;', 'receive output from the trained image classification CNN;', "determine, based on the output, that IPMNs are likely present in the subject's pancreas;", "in response to determining that IPMNs are likely present in the subject's pancreas, cause an indication that IPMNs are likely present in the subject's pancreas to be presented to the user."], 'at least one hardware processor that is programmed to2. The system of claim 1 , further comprising the MRI scanner.3. The system of claim 1 , wherein ...

DIAGNOSIS SUPPORT DEVICE, LEARNING DEVICE, DIAGNOSIS SUPPORT METHOD, LEARNING METHOD, AND PROGRAM

A device for supporting diagnosis has: a reception unit that is configured to receive a fundus image of a subject eye; an identification unit provided with a trained model that is configured to recognize, in the fundus image of the subject eye, an area of abnormality in blood circulation, wherein the model has been trained based upon an image of a fundus and an area of abnormality in blood circulation specified in a fluorescent angiography image of the fundus; and an output unit that is configured to output information relating to the area of abnormality in blood circulation recognized in the fundus image of the subject eye. 110.-. (canceled)11. A method for supporting diagnosis comprising:receiving a fundus image,processing the fundus image using a trained model configured to recognize an area of abnormality in blood circulation in the fundus image, wherein the trained model has been trained based upon a training data set comprising of a training fundus image, a fluorescent angiography image, and information of blood circulation abnormality associated with the fluorescent angiography image, andoutputting information relating to the area of abnormality in blood circulation in the fundus image.12. The method for supporting diagnosis according to claim 11 ,wherein the fluorescent angiography image corresponds to the training fundus image.13. The method for supporting diagnosis according to claim 11 ,wherein the fluorescent angiography image and the training fundus image is acquired from same patient.14. The method for supporting diagnosis according to claim 11 ,wherein the fundus image is the image obtained without performing fundus fluorescein angiography.15. The method for supporting diagnosis according to claim 11 ,wherein the area of abnormality in blood circulation in the fundus image is recognized by specifying a feature common in the training fundus image associated with information of blood circulation abnormality.16. The method for supporting diagnosis ...

Systems and methods for a brain acoustic resonance intracranial pressure monitor

In some aspects, the described systems and methods provide for a method comprising transmitting to a brain of a patient, with at least one transducer, acoustic signals. The method further comprises receiving from the brain, with the at least one transducer, data acquired from the brain including information related to standing waves, distribution of acoustic modes, frequency response, and/or impulse/transient response. The method further comprises determining, from the acquired data, intracranial pressure of the person.

System and methods using real-time predictive virtual 3d eye finite element modeling for simulation of ocular structure biomechanics

Disclosed are systems, devices and methods for performing simulations using a multi-component Finite Element Model (FEM) of ocular structures involved in ocular accommodation.

Systems and methods for tumor detection

In some aspects, the described systems and methods provide for a method comprising transmitting to a brain and/or skull of a patient, with at least one transducer, acoustic signals. The method further comprises receiving from the brain and/or skull, with the at least one transducer, data acquired from the brain and/or skull including information related to standing waves, guided waves, distribution of acoustic modes, frequency response, and/or impulse/transient response. The method further comprises determining, from the acquired data, presence of a tumor within the brain of the person.

ROBOTIC TRAINING APPARATUS AND METHODS

Apparatus and methods for training of robotic devices. Robotic devices may be trained by a user guiding the robot along target trajectory using an input signal. A robotic device may comprise an adaptive controller configured to generate control commands based on one or more of the user guidance, sensory input, and/or performance measure. Training may comprise a plurality of trials. During first trial, the user input may be sufficient to cause the robot to complete the trajectory. During subsequent trials, the user and the robot's controller may collaborate so that user input may be reduced while the robot control may be increased. Individual contributions from the user and the robot controller during training may be may be inadequate (when used exclusively) to complete the task. Upon learning, user's knowledge may be transferred to the robot's controller to enable task execution in absence of subsequent inputs from the user 123.-. (canceled)24. A robotic apparatus , comprising:a platform comprising one or more controllable elements; anda controller communicatively coupled to the platform to operate individual ones of the one or more controllable elements, the controller configured to:receive a first input from a human;analyze the first input and cause the platform to execute a first action in accordance with the first input, wherein the analysis of the first input comprises a determination of a deviation between the first action and a target action; andreceive a second input from a human subsequent to the receipt of the first input, the second input being configured to cause a corrected action by the platform, the corrected action being characterized at least in part by a lower deviation from the target action.25. The robotic apparatus of claim 24 , wherein the first action is based on the operation of the individual ones of the one or more controllable elements of the platform.26. The robotic apparatus of claim 24 , wherein the first input and the second input are ...

WIND VECTOR PREDICTION METHOD USING ARTIFICIAL INTELLIGENCE NETWORK AND ANALYSIS APPARATUS

A wind vector prediction method includes receiving, by an analysis device, a weather video; inputting, by the analysis device, a first base video of a first time point for at least one water vapor absorption channel among a plurality of water vapor absorption channels included in the weather video and a reference video of a time point different from the first time point into a first learning network, and predicting, by the analysis device, a wind vector of a second time point after the first time point for the at least one water vapor absorption channel based on information output from the first learning network. 1. A wind vector prediction method comprising the steps of:receiving, by an analysis device, a weather video;inputting, by the analysis device, a first base video of a first time point for at least one water vapor absorption channel among a plurality of water vapor absorption channels included in the weather video and a reference video of a time point different from the first time point into a first learning network; andpredicting, by the analysis device, a wind vector of a second time point after the first time point for the at least one water vapor absorption channel based on information output from the first learning network,wherein the first learning network is a learning network that outputs video information of the second time point using the first base video and the reference video.2. The method of claim 1 , wherein the reference video is a video of the first base video of a past time point claim 1 , and the first learning network predicts the wind vector based on a difference between the first base video and the reference video claim 1 , or the reference video is a video of the first base video of a future time point claim 1 , and the first learning network predicts the wind vector of a time point between the first base video and the reference video based on the difference between the first base video and the reference video.3. The method of claim 1 ...

A METHOD AND SYSTEM FOR DETERMINING PARAMETERS USED TO MANUFACTURE AN OPTICAL ARTICLE AND A CORRESPONDING OPTICAL ARTICLE

This method for determining specification parameters used to manufacture an optical article includes steps of: training at least one neural network to predict the specification parameters, by using a training data set including a plurality of training prescription parameters and corresponding training specification parameters; and predicting specification parameters of the optical article by the at least one neural network, from prescription parameters relating to the optical article, on the basis of the training. 1. A method for determining specification parameters used to manufacture at least one ophthalmic lens , said specification parameters including optical and/or geometrical data relating to said at least one ophthalmic lens , and/or at least one parameter relating to a manufacturing process required to manufacture said ophthalmic lens , the method comprising steps of:training at least one neural network to predict said specification parameters, by using a training data set comprising a plurality of training prescription parameters and corresponding training specification parameters; andpredicting specification parameters of said at least one ophthalmic lens by means of said at least one neural network, from prescription parameters relating to said at least one ophthalmic lens, on the basis of said training.2. The method according to claim 1 , wherein said prescription parameters relating to said at least one ophthalmic lens comprise at least one parameter relating to an ophthalmic correction for correcting a given visual deficiency.3. The method according to claim 1 , wherein said training step comprises teaching said neural network how to select an appropriate blocking ring for manufacturing said at least one ophthalmic lens.4. The method according to claim 1 , wherein said neural network has 8 input neurons and 24 output neurons.5. The method according to claim 1 , wherein said training step comprises an iterative phase including measuring an error in the ...

Efficiency improvement for machine learning of vehicle control using traffic state estimation

A method of improving efficiency of a vehicle behavior controller using a traffic state estimation network is described. The method includes feeding an input of a feature extraction network of the vehicle behavior controller with a sequence of images. The sequence of images include a highway section and corresponding traffic data. The method also includes disentangling an estimated behavior of a controlled ego vehicle. by the traffic state estimation network. The traffic state estimate network disentangles the estimated of the controlled ego vehicle from extracted traffic state features of the input provided by the feature extraction network. The method further includes selecting an action to adjust an autonomous behavior of the controlled ego vehicle according to the estimated behavior of the controlled ego vehicle.

LANE CHANGE CONTROL DEVICE AND METHOD FOR AUTONOMOUS VEHICLE

A lane change control device and a method for an autonomous vehicle improve safety and accuracy in changing lanes on a road. In particular, the lane change control device includes: a learning device that learns an environment in which the autonomous vehicle is able to change lanes on a road; and a controller that controls a lane change of the autonomous vehicle, based on a learned result of the learning device. 1. A lane change control device of an autonomous vehicle , comprising:a learning device configured to learn an environment in which the autonomous vehicle is able to make a lane change; anda controller configured to control the lane change of the autonomous vehicle, based on a learned result of the learning device.2. The lane change control device of claim 1 , wherein the controller is configured to control the lane change of the autonomous vehicle claim 1 , considering whether a rearward vehicle travelling in a target lane yields to the autonomous vehicle claim 1 , even when it is determined that the autonomous vehicle is able to make the lane change.3. The lane change control device of claim 2 , wherein the controller is configured to stop the autonomous vehicle and re-determine whether the autonomous vehicle is able to make the lane change when the rearward vehicle does not yield to the autonomous vehicle during the lane change of the autonomous vehicle.4. The lane change control device of claim 3 , wherein the controller is configured to determine that the rearward vehicle yields to the autonomous vehicle when a speed of the rearward vehicle is reduced claim 3 , and the controller is configured to determine that the rearward vehicle does not yield to the autonomous vehicle when the speed of the rearward vehicle is maintained or increased.5. The lane change control device of claim 4 , wherein the controller is configured to determine whether the rearward vehicle yields to the autonomous vehicle claim 4 , additionally considering whether signal lamps of the ...

Prediction device, prediction method, computer program product, and vehicle control system

A prediction device includes a lane information obtaining unit, an estimating unit, and a predicting unit. The observation-value-obtaining unit obtains an observation value of mobile object' movement. The lane information obtaining unit obtains lane information regarding lanes within a threshold distance of the mobile object. Based on the observation value and the lane information, the estimating unit estimates temporal change volume and likelihood information. The predicting unit identifies target lanes based on the likelihood information, and for the identified target lanes, based on current-state information indicating one or more states of the mobile object indicated by the observation value obtained at the reference time and based on the temporal change volume, calculates predicted-state information indicating one or more states of the mobile object after the reference time when the mobile object moves to the target lane.

Agent trajectory prediction using anchor trajectories

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for agent trajectory prediction using anchor trajectories.

Prediction control method, input system and computer readable recording medium

A prediction control method is suitable for a display device to display an input moving signal. The prediction control method includes matching a plurality of coordinates corresponding to the input moving signal with a plurality of specific coordinates of the display device and predicting the input moving signal, so that the display device displays the predicted coordinates of the input moving signal.

EVALUATING CHATBOTS FOR KNOWLEDGE GAPS

Knowledge gaps in a chatbot are identified with reference to a domain-specific document and a set of QA pairs of the chatbot. Entities and/or entity values associated with the document are compared to the entities and/or entity values of the QA pairs. Entities of the document not associated with the QA pairs are identified as knowledge gaps. The QA pairs and knowledge gaps are ranked by relevance to the domain. 1. A computer-implemented method comprising:identifying within a document a set of text portions and corresponding document entities;classifying each text portion of a subset of text portions according to a binary classification scheme to establish classified text portions within the set of text portions;associating non-classified text portions with classified text portions according to document entities corresponding to the text portions;classifying the non-classified text portions according to associations with classified text portions;identifying a set of document entities associated with text portions having a first classification;determining a set of chatbot entities associated with question-answer pairs formed by a chatbot; andidentifying gap entities present within the set of document entities and not present within the set of chatbot entities;wherein:the gap entities represent the knowledge gaps of the chatbot.2. The computer-implemented method of claim 1 , further comprising:sorting the set of text portions by corresponding document entities; andselecting the subset of text portions to include each document entity represented in the set of text portions.3. The computer-implemented method of claim 1 , wherein the chatbot has domain knowledge of a first field claim 1 , the domain knowledge being based on a knowledge base4. The computer-implemented method of claim 3 , wherein the knowledge base includes the document.5. The computer-implemented method of claim 1 , further comprising:displaying the identified gap entities and corresponding text portions ...

APPARATUS FOR TRAINING RECOGNITION MODEL, APPARATUS FOR ANALYZING VIDEO, AND APPARATUS FOR PROVIDING VIDEO SEARCH SERVICE

Disclosed herein is an apparatus for analyzing a video shot. The apparatus includes at least one program, memory in which the program is recorded, and a processor for executing the program. The program may include a frame extraction unit for extracting at least one frame from a video shot, a shot composition and camera position recognition unit for predicting shot composition and a camera position for the extracted at least one frame based on a previously trained shot composition recognition model, a place and time information extraction unit for predicting a shot location and a shot time for the extracted at least one frame based on previously trained shot location recognition model and shot time recognition model, and an information combination unit for combining pieces of information, respectively predicted for the at least one frame, for each video shot and tagging the video shot with the combined pieces of information. 1. An apparatus for training a recognition model , comprising:at least one program and memory in which the program is recorded; anda processor for executing the at least one program,wherein the at least one program includesa shot composition recognition model generation unit for generating a neural network model for predicting a shot composition and a camera position using a video shot tagged with shot composition information and camera position information as training data, anda shot time and location recognition model generation unit for generating a neural network model for predicting a shot time and a shot location using a video shot tagged with shot time information and shot location information as training data.2. The apparatus of claim 1 , wherein the shot composition recognition model generation unit comprises:a frame extraction unit for extracting at least one frame from the video shot and forming data for each of the at least one frame;an image feature extraction unit for extracting image features related to an object included in the ...

Method and device for creating a machine learning system

A method for creating a machine learning system which is designed for segmentation and object detection in images. The method includes: providing a directed graph; selecting a path through the graph, at least one additional node being selected from this subset, a path through the graph from the input node along the edges via the additional node up to the output node being selected; creating a machine learning system as a function of the selected path; and training the machine learning system created.

Method and appartaus for data efficient semantic segmentation

A method and system for training a neural network are provided. The method includes receiving an input image, selecting at least one data augmentation method from a pool of data augmentation methods, generating an augmented image by applying the selected at least one data augmentation method to the input image, and generating a mixed image from the input image and the augmented image.

Downsampling genomic sequence data

In an approach to automatically downsampling DNA sequence data using variational autoencoders and preserving genomic integrity of an original file embodiments execute, by an encoder, bootstrapping on genomic sequence data to produce resamples. Furthermore, embodiments assess, by the encoder, unrepresentativeness and self-inconsistency of the resamples and selecting a representative resample according to the assessment, and build, by a modified encoder, vector representations from genotype likelihoods based on the selected representative sample. Additionally, embodiments integrate, by an analytics engine, mapping positional information and the genotype likelihoods to identify an optimum vector representation of a resample, and decode, by a modified decoder, the identified optimum vector representation of the resample to obtain a down-sampled read file that resembles and maintains the genomic integrity of the original file.

IMAGE PROCESSING NEURAL NETWORKS WITH DYNAMIC FILTER ACTIVATION

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing images using neural networks. One of the methods includes receiving a network input; processing the network input through a gater neural network to generate a gating vector that includes a respective value for each of a plurality of filters; determining, from the gating vector and for each of the plurality of filters, whether the filter is active or inactive; and processing the network input through the main convolutional neural network to generate an image processing output, comprising, for each convolutional layer in the first plurality of convolutional layers: receiving an input feature map for the convolutional layer; and generating an output feature map, the generating comprising: for each filter of the convolutional layer that is inactive: setting the output channel for the filter to have all zero elements. 1. A computer-implemented method of processing a network input comprising one or more images through a main convolutional neural network to generate an image processing output for an image processing task , wherein the main convolutional neural network comprises a first plurality of convolutional layers each having a respective plurality of filters , and wherein the method comprises:receiving the network input;processing the network input through a gater neural network, wherein the gater neural network is configured to process the network input to generate a gating vector that includes a respective value for each of the plurality of filters of each of the first plurality of convolutional layers;determining, from the gating vector and for each of the plurality of filters of each of the first plurality of convolutional layers, whether the filter is active or inactive for the processing of the network input; and receiving an input feature map for the convolutional layer; and', [ 'performing a convolution between the input feature map and the filter to ...

ARTIFICIAL NEURAL NETWORK COMPUTATION ACCELERATION APPARATUS FOR DISTRIBUTED PROCESSING, ARTIFICIAL NEURAL NETWORK ACCELERATION SYSTEM USING SAME, AND ARTIFICIAL NEURAL NETWORK ACCELERATION METHOD THEREFOR

An artificial neural network computation acceleration apparatus for distributed processing includes an external main memory for storing input data and synapse weights for input neurons; an internal buffer memory for storing a synapse weight and input data required for each cycle constituting the artificial neural network computation; a DMA module for directly transmitting/receiving data to/from the external main memory and the internal buffer memory; and a general-use communication media block capable of transmitting/receiving the input data and the synapse weights for the input neurons and a result of the computation performed by the neural network computation device to/from another acceleration apparatus physically connected regardless of the type of an integrated circuit. 1. An artificial neural network computation acceleration apparatus for distributed processing to process a computation of an artificial neural network in which input neurons are hierarchically configured , the apparatus comprising:an external main memory configured to store input data and synaptic weights for the input neurons;an internal buffer memory configured to store a synaptic weight and input data required for each cycle constituting the artificial neural network computation among synaptic weights and input data stored in the external main memory;a DMA module configured to directly transmit and receive data to and from the external main memory and the internal buffer memory;a neural network computation device configured to repeatedly process, for each cycle constituting the artificial neural network computation, a series of sequential steps of reading the synapse weight and the input data stored in the internal buffer memory so as to perform an artificial neural network computation and store a computation result in the external main memory;a CPU configured to control an operation of storing the input data and the synapse weights for the input neurons in the external main memory and the ...

CONFIDENCE CLASSIFIERS FOR DIAGNOSTIC TRAINING DATA

Automated assignment of confidence levels to medical diagnoses in a machine learning training data with respect to annotations made upon review of medical records such as x-ray films and test results. Confidence levels support machine learning for computer-aided diagnostic activity. 1. A computer-implemented method for assigning confidence levels to training data for machine learning models comprising:receiving a set of medical records for a medical exam including a corresponding diagnostic annotation for each medical record;identifying a set of diagnostic activity data associated with a first diagnostic annotation of a first medical record;determining a first confidence level for the first diagnostic annotation with reference to the diagnostic activity data; andgenerating a set of training data from the set of medical records with diagnostic annotations and corresponding confidence level assignments including the first medical record, the first diagnostic annotation, and the first confidence level.2. The computer-implemented method of further comprising:weighting the diagnostic annotations of the medical diagnosis according to the determined confidence levels; andwherein:the set of training data includes the weighted diagnostic annotations.3. The computer-implemented method of further comprising:training a convolutional neural network with the weighted diagnoses as part of the modified training data.4. The computer-implemented method of claim 1 , wherein:the set of medical records includes images generated during the medical exam; andthe set of diagnostic activity data for a medical record includes how long an annotator viewed an image of the medical record when annotating the image.5. The computer-implemented method of claim 1 , wherein:the medical exam is a breast cancer screening;the image is a mammogram; and a) no finding,', 'b) benign,', 'c) malignant, and', 'd) suspicious., 'the annotation is a member selected from the group consisting of6. The computer- ...

PREVENTING GLITCH PROPAGATION

When a signal glitches, logic receiving the signal may change in response, thereby charging and/or discharging nodes within the logic and dissipating power. Providing a glitch-free signal may reduce the number of times the nodes are charged and/or discharged, thereby reducing the power dissipation. A technique for eliminating glitches in a signal is to insert a storage element that samples the signal after it is done changing to produce a glitch-free output signal. The storage element is enabled by a “ready” signal having a delay that matches the delay of circuitry generating the signal. The technique prevents the output signal from changing until the final value of the signal is achieved. The output signal changes only once, typically reducing the number of times nodes in the logic receiving the signal are charged and/or discharged so that power dissipation is also reduced. 1. A circuit , comprising:a delay circuit configured to generate a ready signal that is negated at a first transition of a clock signal and asserted after a first delay relative to the first transition, wherein the first delay is at least as long as a second delay; and receive an input signal generated by combinational logic, wherein a change in a first signal received at an input of the combinational logic causes a corresponding change in the input signal at an output of the combinational logic after the second delay following the first transition of a clock signal; and', 'sample the input signal while the ready signal is asserted to transfer a level of the input signal to an output signal of the sampling circuit, wherein the input signal is unchanged from the second delay until the input signal is sampled., 'a sampling circuit configured to2. The circuit of claim 1 , wherein the sampling circuit is further configured to hold the output signal at a constant level from the first transition of the clock signal until the input signal is sampled.3. The circuit of claim 1 , wherein the sampling ...

Automated Construction of Neural Network Architecture with Bayesian Graph Exploration

A system for automated construction of an artificial neural network architecture is provided. The system includes a set of interfaces and data links configured to receive and send signals, wherein the signals include datasets of training data, validation data and testing data, wherein the signals include a set of random number factors in multi-dimensional signals X, wherein part of the random number factors are associated with task labels Y to identify, and nuisance variations S. The system further includes a set of memory banks to store a set of reconfigurable deep neural network (DNN) blocks, hyperparameters, trainable variables, intermediate neuron signals, and temporary computation values including forward-pass signals and backward-pass gradients. The system further includes at least one processor, in connection with the interface and the memory banks, configured to submit the signals and the datasets into the reconfigurable DNN blocks, wherein the at least one processor is configured to execute a Bayesian graph exploration using the Bayes-Ball algorithm to reconfigure the DNN blocks such that redundant links are pruned to be compact by modifying the hyperparameters in the memory banks. The system realizes nuisance-robust variational Bayesian inference to be transferable to new datasets in semi-supervised settings. 1. A system for automated construction of an artificial neural network architecture , comprising:a set of interfaces and data links configured to receive and send signals, wherein the signals include datasets of training data, validation data and testing data, wherein the signals include a set of random number factors in multi-dimensional signals X, wherein part of the random number factors are associated with task labels Y to identify, and nuisance variations S;a set of memory banks to store a set of reconfigurable deep neural network (DNN) blocks, wherein the reconfigurable DNN block is configured either for encoding the multi-dimensional signals X ...

REGULARIZED NEURAL NETWORK ARCHITECTURE SEARCH

A method for receiving training data for training a neural network (NN) to perform a machine learning (ML) task and for determining, using the training data, an optimized NN architecture for performing the ML task is described. Determining the optimized NN architecture includes: maintaining population data comprising, for each candidate architecture in a population of candidate architectures, (i) data defining the candidate architecture, and (ii) data specifying how recently a neural network having the candidate architecture has been trained while determining the optimized neural network architecture; and repeatedly performing multiple operations using each of a plurality of worker computing units to generate a new candidate architecture based on a selected candidate architecture having the best measure of fitness, adding the new candidate architecture to the population, and removing from the population the candidate architecture that was trained least recently. 1. (canceled)2. A method comprising:receiving training data for training a neural network to perform a machine learning task, the training data comprising a plurality of training examples and a respective target output for each of the training examples; and maintaining population data comprising, for each candidate architecture in a population of candidate architectures, (i) data defining the candidate architecture, and (ii) data specifying how recently a neural network having the candidate architecture has been trained while determining the optimized neural network architecture, and', selecting, by the worker computing unit, a plurality of candidate architectures from the population,', 'training, for each selected candidate architecture and by the worker computing unit, a new neural network having the candidate architecture on a training subset of the training data to determine trained values of parameters of the new neural network;', 'determining, for each selected candidate architecture and by the worker ...

TRANSACTION-ENABLED METHODS FOR PROVIDING PROVABLE ACCESS TO A DISTRIBUTED LEDGER WITH A TOKENIZED INSTRUCTION SET

Transaction-enabled methods for providing provable access to a distributed ledger with a tokenized instruction set for polymer production processes are described. A method may include accessing a distributed ledger comprising an instruction set for a polymer production process and tokenizing the instruction set. The method may further include interpreting an instruction set access request and providing a provable access to the instruction set. The method may further include providing commands to a production tool of the polymer production process and recording the transaction on the distributed ledger. 1. A method , comprising:accessing a distributed ledger comprising an instruction set, wherein the instruction set comprises an instruction set for a polymer production process;tokenizing the instruction set;interpreting an instruction set access request;in response to the instruction set access request, providing a provable access to the instruction set;providing commands to a production tool of the polymer production process in response to the instruction set access request; andrecording a transaction on the distributed ledger in response to the providing commands to the production tool.2. The method of claim 1 , wherein the instruction set comprises an instruction set for a chemical synthesis subprocess of the polymer production process.3. The method of claim 2 , further comprising providing commands to a production tool of the chemical synthesis subprocess of the polymer production process in response to the instruction set access request and recording a transaction on the distributed ledger in response to the providing commands to the production tool of the chemical synthesis subprocess of the polymer production process.4. The method of claim 1 , wherein the instruction set comprises a field programmable gate array (FPGA) instruction set.5. The method of claim 1 , wherein the instruction set further includes an application programming interface (API).6. The ...

Short-term load forecasting

A method, computer program, and computer system are provided for load forecasting. Datasets corresponding to source machine learning models and a target domain base model are identified. A set of forecasting models corresponding to the identified datasets are learned. An ensemble model is determined from the learned set of forecasting models based on gradient boosting. An available resource is allocated based on the ensemble model.

NEURAL NETWORK ARCHITECTURE FOR EFFICIENT RESOURCE ALLOCATION

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for efficiently allocating resources among participants. Methods can include receiving valuation data specifying, for each of a plurality of entities, a respective valuation for each of a plurality of resource subsets, each resource subset comprising a different combination of one or more resources of a plurality of resources. After receiving valuation data, assigning each resource in the plurality of resources to a respective entity of the plurality of entities based on the valuations and generating, for each particular entity, a respective input representation that is derived from valuations of every other entity in the plurality of entities other than the particular entity. The input representation for each particular entity is processed using a neural network to generate a rule for the particular entity and a payment based on the rule output for the entities. 1. A method implemented by one or more programmable computers executing one or more instructions stored in one or more storage devices , comprising:receiving, by the one or more programmable computers, valuation data specifying, for each of a plurality of entities, a respective valuation for each of a plurality of resource subsets, each resource subset comprising a different combination of one or more resources of a plurality of resources;assigning, by the one or more programmable computers, each resource in the plurality of resources to a respective entity of the plurality of entities based on the valuations;generating, by the one or more programmable computers, for each particular entity of the plurality of entities, a respective input representation tensor that includes a plurality of channels with each channel being a matrix representing information of the valuations for the plurality of resource subsets from every other entity in the plurality of entities other than the particular entity, the plurality of ...

Unsupervised deformable image registration method using cycle-consistent neural network and apparatus therefor

Disclosed are an unsupervised learning-based image registration method using a neural network with cycle consistency and an apparatus therefor. An image registration method includes receiving a first image and a second image for image registration, outputting a deformation field for the first image and the second image using an unsupervised learning-based neural network with cycle consistency for the deformation field, and generating a registration image for the first image and the second image based on a spatial deformation function using the output deformation field. The outputting of the deformation field includes outputting the deformation field for the first image for registering the first image to the second image may be output, when the first image is a moving image and the second image is a fixed image, and the generating of the registration image includes generating the registration image by applying the deformation field for the first image to the first image using the spatial deformation function.

CROSS-CONTEXT NATURAL LANGUAGE MODEL GENERATION

Provided is a method including obtaining a corpus and an associated set of domain indicators. The method includes learning a set of vectors in an embedding space based on n-grams of the corpus. The method includes updating ontology graphs comprising a set of vertices and edges associating the set of vertices with each other. The method also includes determining a vector cluster using hierarchical clustering based on distances of the set of vectors with respect to each other in the embedding space and determining a hierarchy of the ontology graphs based on a set of domain indicators of a respective set of vertices corresponding to vectors of the vector cluster. The method also includes updating an index based on the ontology graphs. 1. A computer-implemented method of using domain-specific ontologies to of providing summaries of documents in a corpora of natural-language text documents , the method comprising:obtaining, with a computer system, a set of user-specific context parameters and a natural-language text document;determining, with the computer system, a first domain of knowledge based on the set of user-specific context parameters, wherein the first domain of knowledge maps to a first ontology amongst a plurality of ontologies, and wherein ontologies in the plurality of ontologies map n-grams onto a set of concepts to which the n-grams refer;scoring, with the computer system, a first set of n-grams of the natural-language text document using a scoring model based on relations between members of the first set of n-grams;selecting, with the computer system, text sections of the natural-language text based on n-gram scores provided by the scoring model;determining, with the computer system, an initial set of n-grams of the n-grams, wherein each respective n-gram of the initial set of n-grams maps to a respective concept of the set of concepts, and wherein each respective n-gram is identified by an ontology other than the first ontology;determining, with the ...

NEURAL NETWORK LOCALIZATION SYSTEM AND METHOD

A neural network system for inferring a location of a target from a plurality of localization parameters derived from a wireless signal and a method of training thereof. The neural network system comprises first and second neural networks. The plurality of localization parameters comprise one or more parameters relating to a velocity of a target and one or more other parameters. The first neural network is trained to infer a set of candidate locations of a target from values of the one or more other parameters. The second neural network is trained to infer a location of the target from values of the one or more parameters relating to a velocity of the target and a set of candidate locations of the target. 1. A method of training a neural network system to infer a location of a target from values of a plurality of localization parameters derived from a wireless signal from the target ,wherein the neural network system comprising a first neural network and a second neural network,the plurality of localization parameters comprise one or more parameters relating to a velocity of a target and one or more other parameters, and training the first neural network to infer a set of candidate locations of a target from values of the one or more other parameters; and', 'training the second neural network to infer a location of the target from values of the one or more parameters relating to a velocity of the target and a set of candidate locations of the target., 'the method comprises2. A method according to claim 1 , wherein the first neural network is trained using a plurality of first training examples each comprising experimentally determined values of the one or more other parameters derived from a signal emitted by the target at an experimentally measured location claim 1 , and a set of candidate locations derived from the experimentally measured location as an output.3. A method according to claim 2 , wherein the set of candidate locations of each first training example ...

MODEL-ASSISTED DEEP REINFORCEMENT LEARNING BASED SCHEDULING IN WIRELESS NETWORKS

In one embodiment, an apparatus of an access point (AP) node of a network includes an interconnect interface to connect the apparatus to one or more components of the AP node and a processor to: access scheduling requests from a plurality of devices, select a subset of the devices for scheduling of resource blocks in a time slot, and schedule wireless resource blocks in the time slot for the subset of devices using a neural network (NN) trained via deep reinforcement learning (DRL). 1. An apparatus of an access point (AP) node of a network , the apparatus including an interconnect interface to connect the apparatus to one or more components of the AP node , and a processor to:access scheduling requests from a plurality of devices; andselect a subset of the devices for scheduling of resource blocks in a time slot; andschedule wireless resource blocks in the time slot for the subset of devices using a neural network (NN) trained via deep reinforcement learning (DRL).2. The apparatus of claim 1 , wherein the processor is to select the subset of the devices using a random round robin selection.3. The apparatus of claim 1 , wherein the processor is to select the subset of the devices based on a sorting of the subset of devices according to a measure of past throughput for each respective device of the subset.4. The apparatus of claim 3 , wherein the measure of past throughput is a ratio of a channel rate for the device and an exponential weighted average rate previously received by the device.5. The apparatus of claim 1 , wherein the processor is to select the subset of the devices based on a sorting of the subset of devices according to a measure of an amount of data queued for each respective device of the subset.6. The apparatus of claim 5 , wherein the measure of the amount of data queued is one of an instantaneous queue length for the device and a time average of queue length for the device.7. The apparatus of claim 1 , wherein the processor is further claim 1 , for ...

Compositions and methods for diagnosing urinary tract infections

Provided herein are compositions and methods for diagnosing urinary tract infections. In particular, provided herein are compositions and methods for preparing canine urine samples and performing Raman spectroscopy detection of urinary tract infections in the samples.

ARTIFICIAL INTELLIGENCE DEVICE AND ARTIFICIAL INTELLIGENCE SYSTEM FOR MANAGING INDOOR AIR CONDITION

An artificial intelligence (AI) device includes a display, a communication unit configured to transmit indoor air quality information received from an air quality measurement device and meta information input through the display to an AI server, and a processor configured to receive an air quality analysis report generated based on outdoor air quality information, the indoor air quality information and the meta information from the AI server through the communication unit and display the received air quality analysis report on the display. The air quality analysis report includes a first analysis report including a result of analyzing an indoor air quality condition during a measurement period of the air quality measurement device, and a second analysis report including an air quality type according to the indoor air quality condition and a solution for fine dust management according to the air quality type. 1. An artificial intelligence (AI) device , comprising:a display configured to display a meta information screen for inputting meta information about a building;a communication processor configured to:communicate with an air quality measurement device located inside of the building to retrieve indoor air quality information about a quality of air inside the building,transmit the indoor air quality information and the input meta information to an AI server, wherein the AI server acquires outdoor air quality information about a quality of air outside the building; anda processor configured to:control the display to display a first air quality analysis report screen for a particular measurement period of the air quality measurement device measuring the quality of the air inside the building, wherein the first air quality analysis report screen includes an analyzed condition of the indoor air quality based on the indoor air quality information, the meta information and the outdoor air quality information, andcontrol the display to display a second air quality ...

An appliance operation signal processing system and method

An appliance operation signal processing system including an input for receiving an appliance operation signal. The appliance operation signal includes information relating to operation of an appliance. The system further includes an output for outputting information relating to degradation of the appliance and a processor configured to determine a probability that the appliance or a component of the appliance is in at least one degradation state by applying a classifier to a feature vector including information relating to frequency data extracted from the appliance operation signal.

METHOD AND SYSTEM FOR CHARACTERIZING A NANOSTRUCTURE BY MACHINE LEARNING

A method of designing a nanostructure, comprises: receiving a far field optical response and material properties; feeding the synthetic far field optical response and material properties to an artificial neural network having at least three hidden layers; and extracting from the artificial neural network a shape of a nanostructure corresponding to the far field optical response. 1. A method of designing a nanostructure , comprising:receiving a synthetic far field optical response and material properties;feeding said synthetic far field optical response and material properties to an artificial neural network having at least three hidden layers; andextracting from said artificial neural network a shape of a nanostructure corresponding to said far field optical response.2. The method according to claim 1 , wherein said artificial neural network comprises at least two parallel sets of layers claim 1 , wherein said far field optical response and material properties are fed to different sets of layers of said artificial neural network.3. The method according to claim 1 , wherein said far field optical response comprises a spectrum describing a response to a linearly polarized light.4. The method according to claim 1 , wherein said far field optical response comprises a first spectrum describing a response to a horizontally polarized light claim 1 , and a second spectrum describing a response to a vertically polarized light.5. The method according to claim 4 , wherein said artificial neural network comprises three parallel sets of layers claim 4 , wherein said first spectrum is fed to a first set of layers claim 4 , said second spectrum is fed to a second set of layers claim 4 , and said material properties are fed to third set of layers.6. The method according to claim 2 , wherein all parallel sets of layers have the same number of layers.7. The method according to claim 6 , wherein each set of said parallel sets of layers comprises at most five layers.8. The method ...

APPARATUS AND METHOD WITH EGO MOTION ESTIMATION

Disclosed is an ego motion estimation method and apparatus. The ego motion estimation apparatus may generate input data based on radar sensing data collected by one or more radar sensors for each of a plurality of time frames, and estimate ego motion information based on the input data using a motion recognition model. 1. A processor-implemented ego motion estimation method , the method comprising:generating input data based on radar sensing data collected by one or more radar sensors for each of a plurality of time frames; andestimating ego motion information based on the input data using a motion recognition model.2. The method of claim 1 , wherein the estimating of the ego motion information comprises:extracting feature data from the input data using a first model of the motion recognition model; anddetermining the ego motion information based on the feature data using a second model of the motion recognition model.3. The method of claim 1 , wherein the estimating of the ego motion information comprises:determining either one or both of a position and a pose of an apparatus as the ego motion information.4. The method of claim 1 , wherein the estimating of the ego motion information comprises:inputting, as the input data, radar sensing data corresponding to at least two time frames into a layer of the motion recognition model corresponding to one of the at least two time frames.5. The method of claim 1 , wherein the estimating of the ego motion information comprises:extracting current feature data from input data of a current frame and a previous frame of the time frames, using a first model; anddetermining current ego motion information based on the current feature data, using a second model.6. The method of claim 5 , wherein the estimating of the ego motion information comprises:extracting subsequent feature data from input data of a subsequent frame of the time frames and the current frame, using the first model; anddetermining subsequent ego motion information ...

MULTI OPTICALLY-COUPLED CHANNELS MODULE AND RELATED METHODS OF COMPUTATION

An integrated optical module is provided. The optical module comprises multi optically-coupled channels, and enables the use thereof in an Artificial Neural Network (ANN). According to some embodiments the integrated optical module comprises a multi-core optical fiber, wherein the cores are optically coupled. 1. A method of performing a calculation , the method comprising:{'b': 1', '1', '2, 'providing a multi-core optical fiber of a length L comprising at least two cores configured to enable directional light propagation therein along the multi-core optical fiber, the optical fiber is configured to enable evanescent wave coupling between neighboring cores with a coupling length that is shorter than twice the length L at least for light signals having a first wavelength λ and wherein one or more of the cores are amplification core being configured to amplify the λ light according to a power of a control light signal having a second wavelength λ propagating therethrough;'}{'b': '1', 'transmitting input light signals having selected individual powers and the first wavelength λ into a plurality of cores of the multi-core optical fiber;'}obtaining output light signals emitted from one or more of the cores of the multi-core optical fiber, the powers of said output light signals being a function of the powers of the input light signals, and{'b': '2', 'transmitting control light signals having selected individual powers and the second wavelength λ into one or more of the amplification cores of the multi-core optical fiber, thereby defining said function.'}211221. The method of claim 1 , wherein the one amplification core is configured to amplify a λ light—being light at a first wavelength λ propagating therethrough—by a controllable amplification factor determined by a power of a λ light—being light at a second wavelength λ—propagating therethrough simultaneously with the λ light.321. The method of claim 2 , wherein said λ light has a wavelength of about 980 nm and said λ ...

Fault locating method and system based on multi-layer evaluation model

The disclosure discloses a fault locating method based on a multi-layer evaluation model. Firstly, determine a fault type to be inspected and a fault symptom which able to accurately and effectively reflect a power transformer operation status and determine a weight of each fault type by using an association rule and a set pair analysis. Then, establish a DBN model to perform feature extraction and classification on multi-dimensional data of a fault. Finally, perform a comprehensive evaluation on an existing diagnosis result by using the D-S evidence theory. Accordingly, the supporting strength of the common target is reinforced, while the influence of divergent targets is reduced. As a result, the uncertainty in the diagnosis result is significantly reduced. The disclosure is mainly used to monitor and diagnose a status variable of the power transformer in a real-time manner, and treats power transformer status evaluation as a multi-property decision issue.

Method and apparatus for enhancing semantic features of sar image oriented small set of samples

The present disclosure relates to a method for enhancing sematic features of SAR image oriented small set of samples, comprising: acquiring a sample set of an SAR target image, and performing transfer learning and training on the sample set to obtain a initialized deep neural network of an SAR target image, the sample set comprising an SAR target image and an SAR target virtual image; performing network optimization on the deep neural network by an activation function, and extracting features of the SAR target image by the optimized deep neural network to obtain a feature map; and mapping, by an auto-encoder, the feature map between a feature space and a semantic space to obtain a deep visual feature with an enhanced semantic feature.

OPTIMIZATION OF DISCRETE FRACTURE NETWORK (DFN) USING STREAMLINES AND MACHINE LEARNING

A methodology is provided to optimize the dynamic connectivity of a discrete fracture network (DFN) model of a subsurface reservoir against observed reservoir production measures using streamlines and machine learning. Adjustment of discrete fracture network properties of the reservoir is made locally and minimizes computer processing time spent in history matching. An iterative workflow identifies history match issues between measured and predicted or simulated water cut of reservoir produced fluids. Streamline analysis quantifies injector-producer communication and identifies reservoir grid block bundles that dominate dynamic response. A genetic algorithm updates discrete fracture network properties of the reservoir model to improve dynamic history match response. 1. A method of determining a location for drilling a well in a subsurface geological structure of a subsurface hydrocarbon reservoir having existing wells producing fluids comprising hydrocarbons and exhibiting water cut representing water mixed in the fluids being produced , the location being determined indicated by an optimized natural fracture network model of the reservoir , comprising the steps of:(a) obtaining reservoir parameters representing properties of the subsurface reservoir for processing in a data processing system, the reservoir properties including observed cumulative water cut of the produced fluids during production from the existing wells;(b) forming a proposed discrete fracture network model indicating the nature and extent of discrete fractures and fracture flow paths in the reservoir;(c) performing a reservoir simulation history match from the obtained reservoir parameters to determine simulated cumulative water cut of the fluids;(d) determining a measure of the difference between the determined simulated cumulative water cut from the performed reservoir simulation history match and the observed water cut of the produced hydrocarbon fluids from the existing wells;(e) determining ...

POWER GRID AWARE MACHINE LEARNING DEVICE

A system and method for managing operation of electrical devices includes a control module that monitors status of multiple sources of electrical power to one or more electrical devices and electrical usage of the one or more electrical devices that receive electricity from the source of electrical power. The operation of the one or more electrical devices is managed using a machine learning model that forecasts status of the at least one source of electrical power and generates operational rules for the one or more electrical devices from historical values of control parameters of the one or more electrical devices, the status of the source of electrical power, and the electrical usage of the one or more electrical devices. The system may optimize renewable energy utilization, power grid stabilization, cost of electrical power usage, and the like. 1. A computer implemented method of managing operation of electrical devices , comprising:monitoring status of at least one source of electrical power to one or more electrical devices;monitoring electrical usage at each of the one or more electrical devices that receive electricity from the at least one source of electrical power;receiving control parameters for optimization of operation of each of the one or more electrical devices; andmanaging electrical power usage by each of the one or more electrical devices using a machine learning model that forecasts status of the at least one source of electrical power and generates operational rules for each of the one or more electrical devices from historical values of the control parameters, the status of the at least one source of electrical power, and the electrical usage of each of the one or more electrical devices.2. The method of claim 1 , wherein managing electrical power usage comprises generating instructions based on a predictive machine learning model for managing the at least one source of electrical power and the electrical devices claim 1 , the predictive ...

TOUCH PANEL SYSTEM, ELECTRONIC DEVICE, AND SEMICONDUCTOR DEVICE

An electronic device capable of efficiently recognizing a handwritten character is provided. 1. An electronic device comprising:a first circuit;a display portion; anda touch sensor,wherein the first circuit comprises a neural network,wherein the display portion comprises a flexible display,wherein the touch sensor is configured to output an input handwritten character as image information to the first circuit,wherein the first circuit is configured to analyze the image information and convert the image information into character information, and configured to display an image comprising the character information on the display portion, andwherein the analysis is performed by inference through the use of the neural network.2. An electronic device comprising:a first housing;a second housing;a third housing;a plurality of hinges,a first circuit;a display portion; anda touch sensor,wherein the first circuit comprises a neural network,wherein the display portion comprises a flexible display,wherein the flexible display comprises a portion held by the first housing, a portion held by the second housing, and a portion held by the third housing,wherein the touch sensor is configured to output an input handwritten character as image information to the first circuit,wherein the first circuit is configured to analyze the image information and convert the image information into character information, and configured to display an image comprising the character information on the display portion,wherein the analysis is performed by inference through the use of the neural network, andwherein the first housing, the second housing, and the third housing are joined by the plurality of hinges so that the flexible display is changed in shape reversibly between an opened state and a three-folded state.3. The electronic device according to claim 1 , wherein the first circuit comprises a memory capable of retaining analog data.4. The electronic device according to claim 1 , wherein the ...

System and method for continual decoding of brain states to multi-degree-of-freedom control signals in hands free devices

A brain-machine interface system configured to decode neural signals to control a target device includes a sensor to sample the neural signals, and a computer-readable storage medium having software instructions, which, when executed by a processor, cause the processor to transform the neural signals into a common representational space stored in the system, provide the common representational space as a state representation to inform an Actor recurrent neural network policy of the system, generate and evaluate, utilizing a deep recurrent neural network of the system having a generative sequence decoder, predictive sequences of control signals, supply a control signal to the target device to achieve an output of the target device, determine an intrinsic biometric-based reward signal, from the common representational space, based on an expectation of the output of the target device, and supply the intrinsic biometric-based reward signal to a Critic model of the system.

METHOD AND SYSTEM TO ANALYSE PIPELINE CONDITION

A method and system for analysing a condition of a pipeline in real time is disclosed. The method and system comprise generating a transient pressure wave in fluid carried along the pipeline and detecting a transient pressure wave interaction signal responsive to the transient pressure wave where the pressure wave interaction signal has a time duration corresponding to a region of interest of the Generate Transient Pressure pipeline. The method and system further includes processing the transient pressure wave interaction signal to analyse the region of interest of the pipeline. 1. A method for analysing a condition of a pipeline in real time , comprising:generating a transient pressure wave in fluid carried along the pipeline;detecting a transient pressure wave interaction signal responsive to the transient pressure wave, the pressure wave interaction signal having a time duration corresponding to a region of interest of the pipeline; andprocessing the transient pressure wave interaction signal to analyse the region of interest of the pipeline.2. The method of claim 1 , wherein processing the transient pressure wave interaction signal comprises:downsampling in the time domain the transient pressure wave interaction signal to generate a downsampled time window of pressure information;processing the downsampled time window of pressure information by an artificial neural network (ANN) trained to identify a hydraulic feature of a first type and determine associated hydraulic feature characteristics of the hydraulic feature of the first type; andverifying whether the hydraulic feature of the first type occurs in the region of interest of the pipeline.3. The method of claim 2 , wherein verifying whether the hydraulic feature of the first type occurs in the region of interest comprises determining whether one or more of the determined associated hydraulic feature characteristics of the hydraulic feature of the first type are within physical constraints of the pipeline.4. ...

QUOTA MANAGEMENT IN A DATASET MANAGEMENT SYSTEM

Systems and methods for controlling access to a dataset management system using quotas are provided. For example, a request to perform an action in a dataset management system may be obtained from an entity, and a quota record associated with the entity may be selected. Further, it may be determined if the entity has sufficient quota to perform the action. In some examples, when the entity has sufficient quota to perform the action, the request may be allowed. In some examples, when the entity has insufficient quota to perform the action, the request may be denied. 120-. (canceled)21. A system for controlling access to a dataset management system using quotas , the system comprising: receive from an external entity a request to perform an action involving accessing information stored in a dataset management system, the information including a record of annotations;', 'based on an identity of the external entity, select a quota record associated with the external entity of a plurality of quota records;', 'based on the selected quota record, determine whether the external entity has sufficient quota to perform the action;', 'select at least one substitute record of annotations of a plurality of alternative records of annotations based on a similarity between the record of annotations and the at least one substitute record of annotations; and', 'based on determining that the external entity has an insufficient quota, provide an indication to the external entity, the indication including a suggestion of the at least one substitute record of annotations., 'at least one processor configured to22. The system of claim 21 , wherein performing the action comprises training a machine learning algorithm using the information from the dataset management system.23. The system of claim 21 , determining whether the external entity has sufficient quota to perform the action comprises:estimating at least one resource requirement associated with performing the action using the ...

COMPUTER-IMPLEMENTED SIMULATED INTELLIGENCE CAPABILITIES BY NEUROANATOMICALLY-BASED SYSTEM ARCHITECTURE

Computer-implemented systems for simulated intelligence information processing comprising: a digital processing device comprising an operating system configured to perform executable instructions and a memory; a computer program including instructions executable by the digital processing device to create a hierarchical software architecture comprising: a software module for providing a functional interpretation of the prosencephalon, or parts thereof; a software module for providing a functional interpretation of the mesencephalon, or parts thereof; and a software module for providing a functional interpretation of the rhombencephalon, or parts thereof; wherein the software architecture simulates vertebrate, mammalian, primate, or human neuroanatomy. In some embodiments, the systems create simulated intelligence. 139.-. (canceled)40. Non-transitory computer readable media encoded with a computer program including instructions executable by a digital processing device to create a neuroanatomically based software architecture for creation of applications that simulate vertebrate cognitive information processing , the architecture comprising:a. one module configured to implement the functional interpretation of the highest level brain neuroanatomical structure; andb. at least one module configured to implement the functional interpretation of structures within the brain and the connections between said structures;provided that the architecture includes the ability to simulate said functional interpretations of structures and connections; wherein the simulation performs information processing correlating with the cognitive information processing performed by vertebrate, mammalian, primate, or human neuroanatomy.41. The media of claim 40 , wherein the functional interpretation of neuroanatomical structures within the brain is comprised of a hierarchical set of one or more functional interpretations of neuroanatomical structures.42. The media of claim 40 , wherein the ...

Feature extraction using a neurosynaptic system

Embodiments of the invention provide a neurosynaptic system comprising a first set of one or more neurosynaptic core circuits configured to receive input data comprising multiple input regions, and extract a first set of features from the input data. The features of the first set are computed based on different input regions. The system further comprises a second set of one or more neurosynaptic core circuits configured to receive the first set of features, and generate a second set of features by combining the first set of features based on synaptic connectivity information of the second set of core circuits.

CLASSIFYING FEATURES USING A NEUROSYNAPTIC SYSTEM

Embodiments of the invention provide a method comprising receiving a set of features extracted from input data, training a linear classifier based on the set of features extracted, and generating a first matrix using the linear classifier. The first matrix includes multiple dimensions. Each dimension includes multiple elements. Elements of a first dimension correspond to the set of features extracted. Elements of a second dimension correspond to a set of classification labels. The elements of the second dimension are arranged based on one or more synaptic weight arrangements. Each synaptic weight arrangement represents effective synaptic strengths for a classification label of the set of classification labels. The neurosynaptic core circuit is programmed with synaptic connectivity information based on the synaptic weight arrangements. The core circuit is configured to classify one or more objects of interest in the input data 1. A method , comprising:receiving a set of features extracted from input data;training a linear classifier based on the set of features extracted;generating a first matrix using the linear classifier, wherein the first matrix includes multiple dimensions, wherein each dimension includes multiple elements, wherein elements of a first dimension correspond to the set of features extracted, and wherein elements of a second dimension correspond to a set of classification labels;arranging the elements of the second dimension based on one or more synaptic weight arrangements, wherein each synaptic weight arrangement represents effective synaptic strengths for a classification label of the set of classification labels; andprogramming a neurosynaptic core circuit with synaptic connectivity information based on the one or more synaptic weight arrangements, wherein the core circuit is configured to classify one or more objects of interest in the input data.2. The method of claim 1 , wherein the core circuit comprises one or more electronic neurons claim ...

INFORMATION PROCESSING APPARATUS, INFORMATION PROCESSING METHOD, AND NON-TRANSITORY COMPUTER READABLE MEDIUM

An information processing apparatus including inter-class node insertion means for inserting an input vector into a network as an inter-class insertion node. The apparatus further includes a winner node learning time calculation means for incrementing, when an edge is connected between a first winner node and a second winner node, learning time of a node for the first winner node by a predetermined value. The apparatus includes load balancing means for detecting, for each predetermined cycle according to the total number of input vectors, a node where the value of the learning time is relatively large and unbalanced, inserting a new node into a position near the node that has been detected and the adjacent node of the node that has been detected, reducing the learning time of the node that has been detected and the learning time of the adjacent node of the node that has been detected, deleting an edge between the node that has been detected and the adjacent node of the node that has been detected, connecting an edge between the node that has been newly inserted and the node that has been detected, and connecting an edge between the node that has been newly inserted and the adjacent node of the node that has been detected. 1. An information processing apparatus that has a network structure in which nodes described by multidimensional vectors and edges that connect the nodes are arranged and successively receives input vectors which belong to arbitrary classes and learns input distribution structures of the input vectors , the information processing apparatus comprises:inter-class node insertion means for searching a node located at a position that is the closest to the input vector that is input as a first winner node and a node located at a position that is the second closest to the input vector that is input as a second winner node and inserting an inter-class insertion node having the input vector into the network based on a distance between the input vector and ...

FAILURE PREDICTION

A failure prediction system is provided. The system includes a model-based signature generator generating feature vectors from individual attributes of multi-variate time series data based on sequence importance and attribute importance. The system further includes a knowledge database storing feature vectors corresponding to a set of different failure types. The system also includes a set of similarity detectors. Each detect any of the feature vectors generated by the model-based signature generator that are similar to any of the feature vectors corresponding to a respective one of the different failure types stored in the knowledge database based on a similarity threshold and output the respective one of the different failure types and a likely time period when the respective one of the different failure types will occur. 1. A failure prediction system , comprising:a model-based signature generator generating feature vectors from individual attributes of multi-variate time series data based on sequence importance and attribute importance;a knowledge database storing feature vectors corresponding to a set of different failure types; anda set of similarity detectors, each detecting any of the feature vectors generated by the model-based signature generator that are similar to any of the feature vectors corresponding to a respective one of the different failure types stored in the knowledge database based on a similarity threshold and outputting the respective one of the different failure types and a likely time period when the respective one of the different failure types will occur.2. The failure prediction system of claim 1 , wherein the signal generator generates the feature vectors as a matrix.3. The failure prediction system of claim 1 , wherein the signal generator comprises:a feature extractor extracting feature values from the individual attributes of the multi-variate time series and concatenating the feature values into the feature vectors;a feature ...

Modular Polyhedral Computer Architectures and Network Optimization Algorithms

A plurality of processors and routers are mounted on a scalable, modular, polyhedral cluster, creating a mixed hypercube-toroid network. The architecture scales in a lattice model. Therefore within each cluster, the routers are capable of routing messages in hypercube topologies of at least up to six dimensions, and continue by extension to the next cluster on the scaling lattice. Also described herein are various network routing paths derived from one topological embodiment, a cuboctahedron+centroid interconnect, which optimize network traffic for distributed computing, and shared memory applications. Also described herein are mechanical polyhedral scaffoldings for mounting and connecting processors or single board computers. The processor configurations enable function-follows-form computing. Their computing benefits include reduced latency in distributed computing applications, such as swarm movement; improved shared memory; and increased number of interconnects among neighboring nodes, which offers improved neural network computing.

Speech recognition apparatus, speech recognition method, and electronic device

A speech recognition apparatus includes a probability calculator configured to calculate phoneme probabilities of an audio signal using an acoustic model; a candidate set extractor configured to extract a candidate set from a recognition target list; and a result returner configured to return a recognition result of the audio signal based on the calculated phoneme probabilities and the extracted candidate set.

High resolution 3d point clouds generation from downsampled low resolution lidar 3d point clouds and camera images

In one embodiment, a method or system generates a high resolution 3-D point cloud to operate an autonomous driving vehicle (ADV) from a low resolution 3-D point cloud and camera-captured image(s). The system receives a first image captured by a camera for a driving environment. The system receives a second image representing a first depth map of a first point cloud corresponding to the driving environment. The system downsamples the second image by a predetermined scale factor until a resolution of the second image reaches a predetermined threshold. The system generates a second depth map by applying a convolutional neural network (CNN) model to the first image and the downsampled second image, the second depth map having a higher resolution than the first depth map such that the second depth map represents a second point cloud perceiving the driving environment surrounding the ADV.

High resolution 3d point clouds generation based on cnn and crf models

In one embodiment, a method or system generates a high resolution 3-D point cloud to operate an autonomous driving vehicle (ADV) from a low resolution 3-D point cloud and camera-captured image(s). The system receives a first image captured by a camera for a driving environment. The system receives a second image representing a first depth map of a first point cloud corresponding to the driving environment. The system determines a second depth map by applying a convolutional neural network model to the first image. The system generates a third depth map by applying a conditional random fields model to the first image, the second image and the second depth map, the third depth map having a higher resolution than the first depth map such that the third depth map represents a second point cloud perceiving the driving environment surrounding the ADV.

Photonic in-memory co-processor for convolutional operations

A co-processor for performing a matrix multiplication of an input matrix with a data matrix in one step may be provided. The co-processor receives input signals for the input matrix as optical signals. A plurality of photonic memory elements is arranged at crossing points of an optical waveguide crossbar array. The plurality of memory elements is configured to store values of the data matrix. Input signals are connected to input lines of the optical waveguide crossbar array. Output lines of the optical waveguide crossbar array represent a dot-product between a respective column of the optical waveguide crossbar array and the received input signals, and values of elements of the input matrix to be multiplied with the data matrix correspond to light intensities received at input lines of the respective photonic memory elements. Additionally, different wavelengths are used for each column of the input matrix optical signals.

ANALOG MULTIPLY-ACCUMULATE UNIT FOR MULTIBIT IN-MEMORY CELL COMPUTING

Systems, apparatuses and methods include technology that receives, with a first plurality of multipliers of a multiply-accumulator (MAC), first digital signals from a memory array, wherein the first plurality of multipliers includes a plurality of capacitors. The technology further executes, with the first plurality of multipliers, multibit computation operations with the plurality of capacitors based on the first digital signals, and generates, with the first plurality of multipliers, a first analog signal based on the multibit computation operations. 1. A computing system comprising:a processor;a memory array; and receive first digital signals from the memory array,', 'execute multibit computation operations with the plurality of capacitors based on the first digital signals, and', 'generate a first analog signal based on the multibit computation operations., 'a multiply-accumulator (MAC), wherein the MAC includes a first plurality of multipliers that includes a plurality of capacitors, wherein the first plurality of multipliers is configured to2. The computing system of claim 1 , wherein:the plurality of capacitors includes a first group of capacitors and a second group of capacitors, a plurality of switches, and', 'a plurality of branches that include the plurality of switches and the first group of capacitors., 'the first plurality of multipliers further comprises3. The computing system of claim 2 , wherein the second group of capacitors connect the plurality of branches claim 2 , further wherein a capacitance of the second group of capacitors is greater than a capacitance of the first group of capacitors.4. The computing system of claim 2 , wherein the plurality of switches is to be configured to electrically connect or disconnect from an input analog signal based on the first digital signals.5. The computing system of claim 4 , wherein the plurality of capacitors and the plurality of switches form a C-2C ladder.6. The computing system of claim 1 , wherein the ...

MEDIA UNIT RETRIEVAL AND RELATED PROCESSES

Media unit retrieval methods, systems and computer program products are provided that allow a user to search for an item by iteratively presenting media units such as images representing items to the user and receiving user input consisting of selections of the presented media units (including possibly the empty selection). Features, or attributes, a user is interested in, for example semantic features, are inferred from the interaction and media units are retrieved for presentation based on similarity with user-selected media units, through sampling of a probability distribution describing the intent or interests, or combinations of approaches. Accordingly, the user-experience is akin to a conversation about what the user is looking for. Retrieval may be based on both selected and unselected media units and the selection may comprise making a selection with a single action. Further, a database of media units can capture similarity relationships for efficient media unit retrieval. 1. A method of transmitting media units or media unit identifiers for presentation of media units , the method comprisingtransmitting a set of media units or respective media unit identifiers for presentation of the media units, each media unit being associated with a respective feature set defining a set of attribute values for respective attributes or a set of parameters, the parameters defining an attribute-representing probability distribution over attribute values for respective attributes and wherein one or more of the attribute values are derived from neural activations of a hidden layer of an artificial neural network presented with the media unit; and performing the following operations:receiving an input made in response to a presentation of media units of the set of media units;updating an intent probability distribution over attribute values of the attributes using the input;selecting a next set of media units using the updated intent probability distribution; andtransmitting ...

Log-Aided Automatic Query Expansion Based on Model Mapping

Methods, systems, and computer program products for log-aided automatic query expansion based on model mapping are provided herein. A computer-implemented method includes generating a vector representation for each of multiple words derived from historical user queries, wherein each of said vector representations is based on one or more system logs; generating a vector representation for each of multiple documents in a corpus of documents related to solutions to one or more hardware problems and/or one or more software problems; generating a vector representation for a user query based on the generated vector representation for each of the multiple words derived from the historical user queries; comparing the vector representation for the user query to the vector representation for each of multiple documents in the corpus; and determining one or more documents from the corpus to output in response to the user query based on said comparing.

Method for making music recommendations and related computing device, and medium thereof

This application discloses a method for making music recommendations. The method for making music recommendations is performed by a server device. The method includes obtaining a material for which background music is to be added; determining at least one visual semantic tag of the material, the at least one visual semantic tag describing at least one characteristic of the material; identifying a matched music matching the at least one visual semantic tag from a candidate music library; sorting the matched music according to user assessing information of a user corresponding to the material; screening the matched music based on a sorting result and according to a preset music screening condition; and recommending matched music obtained through the screening as candidate music of the material.

Toxic vector mapping across languages

Methods, systems, and devices for language mapping are described. Some machine learning models may be trained to support multiple languages. However, word embedding alignments may be too general to accurately capture the meaning of certain words when mapping different languages into a single reference vector space. To improve the accuracy of vector mapping, a system may implement a supervised learning layer to refine the cross-lingual alignment of particular vectors corresponding to a vocabulary of interest (e.g., toxic language). This supervised learning layer may be trained using a dictionary of toxic words or phrases across the different supported languages in order to learn how to weight an initial vector alignment to more accurately map the meanings behind insults, threats, or other toxic words or phrases between languages. The vector output from this weighted mapping can be sent to supervised models, trained on the reference vector space, to determine toxicity scores.

MACHINE LEARNING MODEL WITH WATERMARKED WEIGHTS

In some examples, a system includes storage storing a machine learning model, wherein the machine learning model comprises a plurality of layers comprising multiple weights. The system also includes a processing unit coupled to the storage and operable to group the weights in each layer into a plurality of partitions; determine a number of least significant bits to be used for watermarking in each of the plurality of partitions; insert one or more watermark bits into the determined least significant bits for each of the plurality of partitions; and scramble one or more of the weight bits to produce watermarked and scrambled weights. The system also includes an output device to provide the watermarked and scrambled weights to another device. 1. A system comprising:a processing unit; receive a machine learning model comprising a plurality of layers, respective ones of the layers comprising multiple weights;', 'determine an accuracy bias for each of multiple different sets of possible values for Np and Nb, wherein an Np of a respective layer is a number of partitions into which to group the weights in the respective layer, and an Nb of a respective partition is a number of least significant bits (LSBs) of the respective partition to be used for watermarking;', 'determine an Np for each of the layers and an Nb for each of the partitions in response to the determined accuracy biases;', 'insert one or more watermark bits into the Nb LSBs of the weights in each of the Np respective partitions in each of the respective layers; and', 'scramble one or more of the weight bits to produce watermarked and scrambled weights; and, 'a memory storing software instructions that, when executed by the processing unit, cause the processing unit toan output device configured to provide the watermarked and scrambled weights to another device.2. The system of claim 1 ,wherein the processing unit is configured to copy the one or more watermark bits from the watermarked and scrambled weights ...

Auto-tuning of rule weights in profiles

Disclosed is a system to optimize rule weights for classifying access requests so as to manage rates of false positives and false negative classifications. A rules suggestion engine may suggest a profile of classification rules to a merchant for access requests. The system can optimize weights for the profile of rules using a cost function based on a training set of historical access requests, for example using stepwise regression or machine learning (ML). The system can compute a profile score based on the optimized weights, for example by summing the weights. The system statistically analyzes the profile score using classification thresholds and the historical access requests. The system can perform receiver operating characteristic (ROC) analysis for various threshold values, enabling a user to select a suitable threshold. The system can further optimize by adding or removing rules from the profile of rules.