Output Coding and Modularity for Multi-Class Problems

Abstract

We study backpropagation networks learning classification problems with multiple classes k > 3. The common way to code the output of a network is the one-per-class (OPC) method, where one bit is assigned to each class. A technique called error-correcting output coding (ECOC) converts the k-class learning problem into a large number of two-class learning problems. We propose to use modular architectures as a way to decorrelate the (redundant) network outputs. Various modular architectures are tested on an artificial problem. We conclude that ECOC only improves upon OPC when combined with a sufficiently modular approach.

Real World Computing Partnership

Foundation for Neural Networks