A study on chaotic dynamics of deep artificial neural network activated by biological neuron model

This paper analyzes the effects of the chaotic signals used by the brain to perform some cognitive functions on the Spiking Neural Network (SNN), defined as the third-generation Artificial Neural Network (ANN) that best represents the biological neuron. In the first phase of the paper, neural networks with different layers are designed to perform classifications like ANN and SNN. Classification performances of these deep networks using the Rectified Linear Unit activation function in ANN mode and the Izhikevich Neuron model in SNN mode are presented comparatively. It is observed that SNNs perform at least as well as ANNs under normal conditions. In the second stage of the study, the classification performances of these deep networks in the SNN mode were analyzed in different chaotic environments, and the findings were reported. In light of the findings, it is seen that SNNs can exhibit a classification success similar to ANNs and maintain this success rate up to a certain chaotic current intensity. Moreover, some levels of chaotic current contribute to the network's classification performance. This is the first study to investigate the chaotic environment behavior of SNNs.