Optimization of recursive branching network

Recursive branching network (RBN) was proposed in [1] to solve linearly non-separable problems using output-coded perceptrons. It relies on splitting the training patterns, at random, between parallel perceptrons. However, the random splitting mechanism can trap the perceptron in conflicting patterns. Optimized splitting methods are proposed here to insure meaningful way of splitting. We propose three splitting methods which use different similarity measures between patterns. We examine these methods on five standard data sets. In general, these methods enhance the performance of RBN and in many cases contribute to lowering the network complexity