Spike Neural Network Learning Algorithm Based on an Evolutionary Membrane Algorithm

Chuang Liu,Wanghui Shen,Yingkui Du,Zhonghu Yuan,Le Zhang

doi:10.1109/access.2021.3053280

Chuang Liu, Wanghui Shen + Show 3 more

Open Access

https://doi.org/10.1109/access.2021.3053280

Copy DOI

Abstract

As one of the important artificial intelligence fields, brain-like computing attempts to give machines a higher intelligence level by studying and simulating the cognitive principles of the human brain. Compared with the traditional neural network, the spiking neural network (SNN) has better biogenesis and stronger computing power. In this paper, an SNN learning model based on an evolutionary membrane algorithm is proposed to solve the problem of supervised classification. The proposed algorithm uses the P system’s object, reaction rules, and membrane structure to solve these problems. Specifically, the proposed algorithm can automatically adjust the learning parameters of the network by adjusting the synaptic weight in the learning stage of the spiking neural model according to different application data, providing a better solution model for balance exploration and exploitation. In the simulation experiment, effectiveness verification research is carried out. The simulation results show that compared with other experimental algorithms, the proposed algorithm has a competitive advantage in solving twelve supervised classification benchmark problems through learning curves and quantified classification results.

Highlights

Spiking neural networks (SNNs) are often referred to as third-generation artificial neural networks, which are generalized approximate and parallel distributed processing models [1]
We use the convergence of different colors to represent the learning curve results of the experimental algorithms
The red line represents the learning curve of the proposed algorithm, the blue line represents the differential evolution (DE) algorithm, the green line represents the particle swarm optimization (PSO) algorithm, the black line represents the cuckoo search (CS) algorithm, the purple line represents the harmony search (HS) algorithm, and the blue and green lines represent the artificial bee colony (ABC) algorithm. It can be seen from the learning curve that the proposed algorithm is superior to these experimental algorithms in 12 supervised classification datasets, which is related to the membrane structure and reaction rules

Summary

Introduction

Spiking neural networks (SNNs) are often referred to as third-generation artificial neural networks, which are generalized approximate and parallel distributed processing models [1]. Different from the traditional perceptron neural network, SNN uses the spike ignition sequence for information processing and uses the spiking time coding method to encode the input variable as the spiking ignition time. The neuronal simulation of SNNs is closer to reality, and the influence of time information is considered. The SNN neurons are not activated in each propagation iteration, and they are only activated when their membrane potential reaches a certain value. When a neuron is activated, it sends signals to other neurons to increase or decrease its membrane potential.

Objectives

Results

Conclusion