ConvSNP: a deep learning model embedded with SNP-like neurons

Inspired from spiking mechanisms in spiking neural P (SNP) systems, this paper proposes a new type of neurons, termed as SNP-like neurons. The mathematical model for SNP-like neurons is a generalized linear function. Based on SNP-like neurons, a new class of deep learning models are developed, called ConvSNP models. By referring the structures of the existing convolutional neural networks (CNNs), five ConvSNP models are designed. The five ConvSNP models are evaluated on three benchmark data sets and compared with the corresponding CNNs. The comparison results demonstrate the availability and effectiveness of ConvSNP models for three classical classification tasks.

Automated summary

Inspired by spiking mechanisms in spiking neural P (SNP) systems, this paper proposes a new type of neurons, termed as SNP-like neurons. Based on SNP-like neurons, a new class of deep learning models called ConvSNP models are developed. Five ConvSNP models are designed by referring to the structures of existing convolutional neural networks (CNNs). The evaluation results on three benchmark data sets demonstrate the availability and effectiveness of ConvSNP models for classical classification tasks.