Evolving Spiking Neural Network as a Classifier: An Experimental Review

Abstract

AbstractThe brain-inspired Spiking Neural Networks (SNNs) are considered as the third generation of neural networks for AI applications. The spiking neural network has been proved very efficient to predict and classify data with spatial and temporal information in the way of modeling the behavior and learning potential of the brain. The aim is to understand the working of the Evolving SNN (ESNN) as a classifier and how it is different from the existing neural models. Besides exploring the existing ESNN architecture, the results have been generated by tuning the various parameters of the ESNN model which may be contributing to provide better prediction accuracies. The tuned ESNN model is applied to various datasets and compared with the existing second-generation neural network model like LSTM. The results show comparable improvement in the classification accuracy using ESNN which concludes that the ESNN and its variants are the beginning of a new era of Neural Networks.KeywordsEvolving spiking neural networkReservoirGaussian receptive fieldSpatial-temporal data