Accuracy Investigation of a Neuromorphic Machine Learning System Due to Electromagnetic Noises Using PEEC Model

Abstract

The accuracy of a neuromorphic machine learning system was investigated using the partial equivalent element circuit (PEEC) model to analyze electromagnetic effects. A multilayer neural network (MNN) model for a classification task was introduced, and the corresponding neuromorphic circuit was designed. An efficient PEEC model for crossbar structures was proposed and validated by comparison with HFSS simulations. The designed neuromorphic circuit including the PEEC crossbar array model was simulated using SPICE while varying operation speed, structure size, and the activation function. Test cases compared electromagnetic noises and the accuracy of the neuromorphic system for the classification task.