The circuital design of generalized cellular automata for parallel optimization

The generalized cellular automata (GCA) has the pyramid architecture and the multi-granularity cellular dynamics for effectively solving a class of optimizations problems. In order to further take advantages of GCA, this paper discusses the hardware implementation of GCA with VLSI systolic techniques. In comparison with the Hopfield-type neural networks and cellular neural networks, the implementation scheme of GCA has features in terms of the much less number of interconnections, the higher-degree optimality, the quicker convergence speed, and the much easier selection of circuital parameters.