AI summary

This paper introduces a supervised learning procedure using mixtures of expert networks, where each network learns a subset of training cases. The model is evaluated on a speaker-independent vowel discrimination problem using formant data from 75 speakers, demonstrating effective task decomposition.

Abstract

We present a new supervised learning procedure for systems composed of many separate networks, each of which learns to handle a subset of the complete set of training cases. The new procedure can be viewed either as a modular version of a multilayer supervised network, or as an associative version of competitive learning. It therefore provides a new link between these two apparently different approaches. We demonstrate that the learning procedure divides up a vowel discrimination task into appropriate subtasks, each of which can be solved by a very simple expert network.