We present a parallel algorithm for laterally interconnected synergetically self-organizing map (LISSOM) neural network, a self-organizing map with lateral excitatory and inhibitory connections, to enhance the computational efficiency. A general strategy of balancing workload for different sizes of LISSOM networks on parallel computers is described. The parallel algorithm of LISSOM is implemented on IBM SP2 and PC cluster. The results demonstrate the efficiency of this LISSOM parallel algorithm in processing networks with large sizes. Parallel implementation for different input dimensions in networks of the same size (i.e., 20×20) show that the speedup can sustain at a high level. We demonstrate the LISSOM can be applied to complex problems through the parallel algorithm devised in this study.
