A model for neuro-computing

Abstract

The authors propose a new model for neural computation based solely on a pattern recognition mechanism. By using pattern recognition as the basis of neurocomputing, the authors demonstrate that such a model can solve nonnumeric problems quickly, but it becomes very slow in solving numeric problems. Decisions in problem solving are also made through pattern matching. The outcome of matching is weighted to reflect appropriately the neurological basis. The complexity of computation is defined in terms of pattern matching operations.