Image classification method based on time domain coding and pulse neural network

Abstract

The invention discloses an image classification method based on time domain coding and a pulse neural network. The method comprises the following steps: S1, constructing a sample set based on time domain coding and category marking of an image; s2, constructing a pulse neural network as a classification model; s3, training the spiking neural network by using the constructed sample set to obtain a maturely trained spiking neural network; and S4, carrying out time domain coding on an image to be identified, and inputting the image to the maturely trained pulse neural network to obtain a classification result of the image. According to the invention, through a direct training framework which does not need to calculate neuron membrane potential, the training difficulty of the spiking neural network is reduced, and then real-time low-power-consumption image recognition and classification are effectively realized.

Description

technical field [0001] The present invention relates to image classification, in particular to an image classification method based on time domain coding and impulse neural network. Background technique [0002] Spiking neural networks (SNNs) have strong biological plausibility, and neurons communicate through spikes, just like biological neurons. They work asynchronously, i.e., generate output spikes without waiting for all input neuron spikes, which brings advantages such as spike sparsity, low latency, and high energy efficiency. [0003] But the performance of SNN has lagged far behind the traditional deep neural network (DNN), one of the main reasons is that SNN is difficult to train. , the DNN is represented by the standard layer response y = f(xW + b), where gradient backpropagation can be efficiently performed. , in contrast, for SNNs, we have to simulate time-domain neuronal membrane potentials with indistinguishable spikes, and gradient descent is difficult and t...

AI Technical Summary

Problems solved by technology

, in contrast, for SNNs, we have to simulate time-domain neuronal membrane potentials with indistinguishable spikes, and gradient descent is difficult and time-consuming. So far, direct training of SNNs has been limited to shallow networks, and no one has directly studied them like ImageNet. SNN is trained on a large data set; when classifying images, it is difficult to classify and recognize images through the spiking neural network because the spiking neural network cannot directly input image information

Images