Learning by supervised clustering and matching

This article presents a procedure for a class of neural networks, known as neural logic networks, to learn multidimensional mapping of both binary and analog data. The procedure, termed supervised clustering and matching (SCM), provides a means of deducing inductive knowledge from training cases. In contrast to gradient descent error correction methods, pattern mapping is learned by fast and incremental clustering of input and output patterns. Specifically, learning/encoding only takes place when both the input and output match criteria are satisfied in a template matching process. To handle sparse and/or noisy data sets, the authors also present a weighted voting scheme whereby distributed cluster activities combine to produce a final output. The performance of the SCM algorithm, compared with alternative systems, is illustrated on a sonar return signal recognition and a sunspot time series prediction problems