PATTERN-DISCRIMINATION USING FEEDFORWARD NETWORKS - A BENCHMARK STUDYOF SCALING BEHAVIOR

The discrimination powers of multilayer perceptron (MLP) and learning vector quantization (LVQ) networks are compared for overlapping gaussi an distributions. It is shown, both analytically and with Monte Carlo studies, that the MLP network handles high-dimensional problem in a mo re efficient way than LVQ. This is mainly due to the sigmoidal form of the MLP transfer function, but also to the fact that the MLP uses hyp erplanes more efficiently. Both algorithms are equally robust to limit ed training sets and the learning curves fall off like 1/M, where M is the training set size, which is compared to theoretical predictions f rom statistical estimates and Vapnik-Chervonenkis bounds.