MAXIMIZING THE ROBUSTNESS OF A LINEAR THRESHOLD CLASSIFIER WITH DISCRETE WEIGHTS

Quantization of the parameters of a perceptron is a central problem in hardware implementation of neural networks using a numerical technolo gy, An interesting property of neural networks used as classifiers is their ability to provide some robustness on input noise. This paper pr esents efficient learning algorithms for the maximization of the robus tness of a perceptron and especially designed to tackle the combinator ial problem arising from the discrete weights.