Supervised Learning Algorithm for Multilayer Spiking Neural Networks with Long-Term Memory Spike Response Model.

Xianghong Lin,Xiangwen Wang,Mengwei Zhang,Thippa Reddy G

doi:10.1155/2021/8592824

Xianghong Lin, Xiangwen Wang + Show 2 more

Open Access

https://doi.org/10.1155/2021/8592824

Copy DOI

Abstract

As a new brain-inspired computational model of artificial neural networks, spiking neural networks transmit and process information via precisely timed spike trains. Constructing efficient learning methods is a significant research field in spiking neural networks. In this paper, we present a supervised learning algorithm for multilayer feedforward spiking neural networks; all neurons can fire multiple spikes in all layers. The feedforward network consists of spiking neurons governed by biologically plausible long-term memory spike response model, in which the effect of earlier spikes on the refractoriness is not neglected to incorporate adaptation effects. The gradient descent method is employed to derive synaptic weight updating rule for learning spike trains. The proposed algorithm is tested and verified on spatiotemporal pattern learning problems, including a set of spike train learning tasks and nonlinear pattern classification problems on four UCI datasets. Simulation results indicate that the proposed algorithm can improve learning accuracy in comparison with other supervised learning algorithms.

Highlights

Research in neuroscience has shown that the precise timing of spikes is used to represent, transmit, and process information in the biological nervous system.e temporal encoding strategy can integrate many aspects of neural information, such as time, space, frequency, phase, etc. [1, 2]. e spiking neuron model is the basic computational units of spiking neural networks (SNNs), where precisely timed spikes are used to transmit the neural information [3, 4]
In order to simulate the activity of brain, it is necessary to study spike train supervised learning algorithm for multilayer feedforward SNNs. e design of spike train learning algorithm is more difficult than that of the single-spike learning for SNNs, but it has more powerful learning capability for solving the complex problems. erefore, in this paper, combining with the longterm memory characteristic of spike response model (SRM), we extend the algorithm proposed by Booij and tat Nguyen [17] and propose a supervised learning algorithm based on the error backpropagation of multiple spikes, which can achieve spike train learning for multilayer feedforward SNNs
65.1% ± 4.7 60.9% ± 3.2 61.5% ± 1.4 60.8% ± 2.0 64.7% ± 2.7 datasets refers to the results described in [45]. e classification accuracy of multi-remote supervised method (ReSuMe), multi-STIP, and our method for the datasets of Iris, Pima Indians Diabetes (PIMA), Wisconsin Breast Cancer (WBC), and Liver is obtained by multilayer feedforward SNNs with long-term memory SRM

Summary

Introduction

Research in neuroscience has shown that the precise timing of spikes (temporal encoding) is used to represent, transmit, and process information in the biological nervous system. Booij and tat Nguyen [17] proposed a supervised learning algorithm for multilayer feedforward SNNs with long-term memory SRM. In this algorithm, neurons in the output layer are still limited to learn single spike. Erefore, in this paper, combining with the longterm memory characteristic of SRM, we extend the algorithm proposed by Booij and tat Nguyen [17] and propose a supervised learning algorithm based on the error backpropagation of multiple spikes, which can achieve spike train learning for multilayer feedforward SNNs. e rest of this paper is organized as follows.

Related Works

Neuron Model and Network Architecture

Supervised Learning Rules Based on Gradient Descent

Analysis and Discussion

Findings

Conclusions