DILUTION IN BOOLEAN PERCEPTRONS THAT LEARN FROM NOISY EXAMPLES

We investigate the effect of dilution after learning on the generaliza tion ability of single-layer Boolean perceptrons that learn from noisy examples. We present a thorough comparison between the relative perfo rmances of several well known learning rules. In particular, we show t hat the effect of dilution is always deleterious, and that the Bayes a lgorithm always gives the best generalization performance.