25-07-2023 дата публикации
Номер: CN116488970A
Автор:
ZHANG QI,
XIN XIANGJUN,
YAO HAIPENG,
YANG RUI,
JIANG JINKUN,
GAO RAN,
TIAN QINGHUA,
TIAN FENG,
WANG YONGJUN,
YANG LEIJING,
LI ZHIPEI,
WANG FU,
WANG XISHUO,
CHANG HUAN,
GUO DONG
Принадлежит:
The invention provides a high-speed optical communication system equalization method, which comprises the following steps of: constructing a neural network equalizer based on Volterra series, enhancing a combined neural network by adopting cross-correlation learning, and combining linear compensation and nonlinear compensation, so that the network does not need an additional linear compensation module any more, and the overall compensation capability of the network is enhanced; through an additional enhanced cross-correlation learning layer, the neural network can better learn second-order and third-order cross-correlation items of an input signal, and the compensation capability of the neural network to system nonlinearity is enhanced. Learning a high-order cross-correlation term of the input signal from the low-order self-correlation input signal through a high-order full-connection layer; and meanwhile, a clustering algorithm is used at the dense weight part in the network, so that the ...
Подробнее