Mathematical simulation of memristive for classification in machine learning

Abstract

Over the last few years, neuromorphic computation has been a widely researched topic. One of the neuromorphic computation elements is the memristor. The memristor is a high density, analogue memory storage, and compliance with Ohm's law for minor potential changes. Memristive behaviour imitates synaptic behaviour. It is a nanotechnology that can reduce power consumption, improve synaptic modeling, and reduce data transmission processes. The purpose of this paper is to investigate a customized mathematical model for machine learning algorithms. This model uses a computing paradigm that differs from standard Von-Neumann architectures, and it has the potential to reduce power consumption and increasing performance while doing specialized jobs when compared to regular computers. Classification is one of the most interesting fields in machine learning to classify features patterns by using a specific algorithm. In this study, a classifier based memristive is used with an adaptive spike encoder for input data. We run this algorithm based on Anti-Hebbian and Hebbian learning rules. These investigations employed two of datasets, including breast cancer Wisconsin and Gaussian mixture model datasets. The results indicate that the performance of our algorithm that has been used based on memristive is reasonably close to the optimal solution.