Notes on Margin Training and Margin p-Values for Deep Neural Network Classifiers

TL;DR

This paper introduces a local class-purity theorem for Lipschitz continuous deep neural networks, discusses margin achievement during training, and presents a method to compute margin p-values for test samples.

Contribution

It presents a new local class-purity theorem, methods for margin training, and a way to compute margin p-values for DNN classifiers.

Findings

New local class-purity theorem for Lipschitz DNNs

Methodology for achieving classification margin during training

Procedure for computing margin p-values for test samples

Abstract

We provide a new local class-purity theorem for Lipschitz continuous DNN classifiers. In addition, we discuss how to achieve classification margin for training samples. Finally, we describe how to compute margin p-values for test samples.