COMPLEX-VALUED TIMING SIGNAL PREDICTION METHOD BASED ON COMPLEX-VALUED NEURAL NETWORK

WIPO Patent Application WO/2023/216383

A1

The present invention relates to a complex-valued timing signal prediction method based on a complex-valued neural network. The method comprises the following steps: constructing a data set of a complex-valued timing signal, and dividing the data set into a training set and a test set; constructing a fully complex neural network, and initializing a parameter; training the fully complex neural network by using the training set and an adaptive orthogonal complex-valued gradient descent method, and stopping training when a preset number of iterations is reached, so as to obtain a trained fully complex neural network; and inputting the test set into the trained fully complex neural network, so as to obtain a predicted value of the complex-valued timing signal. In the present invention, the complex-valued timing signal is directly predicted by using the fully complex neural network, and the adaptive orthogonal complex-valued gradient descent method is used during the training of the fully complex neural network, thereby preventing an intrinsic relationship between complex-valued signals from being destroyed, increasing the convergence speed of training, and improving the prediction precision of a complex-valued timing signal.

HUANG HE (CN)
ZHAO WEIJING (CN)

PCT/CN2022/101004

November 16, 2023

June 24, 2022

Click for automatic bibliography generation  

UNIV SOOCHOW (CN)

G06N3/04; G06N3/08

CN106875002A	2017-06-20
CN113408726A	2021-09-17
CN112182961A	2021-01-05
CN112861813A	2021-05-28
CN113158582A	2021-07-23
US20190251421A1	2019-08-15

CENTRAL SOUTH WELL INTELLECTUAL PROPERTY OFFICE (CN)

View/Download PDF      PDF Help

Previous Patent: POWER GRID FAULT POSITIONING METHOD BASED ON BEIDOU SATELLITE

Next Patent: BATTERY AND ELECTRIC DEVICE