Multilayer Perceptrons: An Introduction

Abstract Deep neural networks have been widely and successfully employed in various fields, such as computer vision, natural language processing, and healthcare. An important and fundamental building block for these deep neural networks is multilayer perceptrons (MLPs). This article reviews the mathematical definition of an MLP and describes the algorithm for estimating its parameters. It also applies an MLP to a well‐known handwritten digit dataset and discusses various factors one should consider when using MLPs.