Competitive Nonlinear Layered Spiking Neural P System for solving classification problems

Abstract

Spiking neural P systems (SN P systems) are a class of membrane computing models that abstract the mechanism of spiking neurons. SN P system has been used in various engineering applications. The flexible structure of the SN P system allows it to be used for designing machine learning algorithms without the need for overly simplified neurons as in neural networks. In this paper, a novel SN P system called competitive nonlinear layered spiking neural P system (CNLSN P system) is proposed for solving classification problems. Experiments on benchmark datasets show that the recognition accuracy of the CNLSN P system is improved by 1.5%–2.5% compared to the layered spiking neural P system (LSN P system). Based on the CNLSN P system, a new class of deep learning model called the ConvCNLSNP model is developed. Experiments on benchmark datasets show that the ConvCNLSNP model reduces the time consumption by 85%–98% while maintaining recognition accuracy comparable to CNNs.