Maximizing the robustness of a linear threshold classifier with discrete weights

Abstract

Quantization of the parameters of a perceptron is a central problem in hardware implementation of neural networks using a numerical technology. An interesting property of neural networks used as classifiers is their ability to provide some robustness on input noise. This paper presents efficient learning algorithms for the maximization of the robustness of a perceptron and especially designed to tackle the combinatorial problem arising from the discrete weights.