International Symposium on Mathematical Science at Niigata University 2016
Generation for Pattern Classification using CA
*Gil-Tak Kong (Pukyong National University (PKNU))
The important prerequisites of designing pattern classifier are high throughput and low cost hardware implementation. The simple, regular, modular and cascadable local neighborhood sparse network of Cellular Automata (CA) suits ideally for low cost VLSI implementation. Thus CA which has the multiple attractor is adapted for use as a pattern classifier. By concatenating two predecessor multiple attractor CA (TPMACA) we can construct a pattern classifier. In this paper we propose a method for finding dependency vector by using state transition matrix of the CA. Also we propose various methods for generating TPMACA corresponding to a given dependency vector.