Mapping method from artificial neural network to spiking neural network

Abstract

The invention relates to the technical field of neural networks, in particular to a mapping method from an artificial neural network to a spiking neural network. The method comprises the following steps: performing unstructured pruning on the artificial neural network, acquiring a random sparse weight matrix, setting a plurality of quantization bits, initializing a centroid by using K-means prototype clustering, determining a quantization threshold value, testing the recognition rate of the artificial neural network, performing fine adjustment on the centroid when the expected precision is not met, and performing comparison to obtain an optimal quantization bit; repeating the quantization process by using the optimal quantization bit, setting a pulse quantization maximum value, and evaluating the sequence similarity before and after quantization by using cross entropy; after the maximum value of the minimum cross entropy is obtained, completing pulsation of the spiking neural network, and finally realizing mapping from the artificial neural network to the spiking neural network. Compared with the prior art, the training process is simple, and the precision lost in the mapping process is low.

Description

technical field [0001] The invention relates to the technical field of neural networks, in particular to a mapping method from an artificial neural network to a pulse neural network. Background technique [0002] Artificial Neural Network (ANN) is a non-linear, self-adaptive information processing system composed of a large number of interconnected processing units. Each node represents a specific output function, called the activation function. Each connection between two nodes represents a weighted value for the signal passing through the connection, called weight, which is equivalent to the memory of the neural network. The output of the network varies according to the way the network is connected, the weight values ​​and the activation function. At present, artificial neural networks are widely used in image processing, natural language processing and other fields. With mature training algorithms, the model performance is better and the performance ability is strong. ...

AI Technical Summary

Problems solved by technology

[0004] In view of the above existing problems or deficiencies, in order to solve the problems of difficult training process and low precision in the existing artificial neural network to pulse neural network method, the present invention provides a mapping method from artificial neural network to pulse neural network

Images