A parallel planar lattice architecture for competitively inhibited neural network simulation

Competitively inhibited neural network (CINN) appears to be promising in adaptive parameter estimation and pattern classification for its intelligent information processing potential. A neurocomputer of the parallel planar lattice architecture (PPLA) based on transputers for large scale, high speed simulations of the CINN is proposed in this paper. It has an even greater expandability of parallelism and exhibits sufficient flexibility to simulate the CINN of various network configurations and parameters. The experiments demonstrate that the performance of this neurocomputer is almost proportional to the number of processors by using a load balancing algorithm.<>