Performance evaluation of bidirectional associative memory on a transputer-based parallel system

Abstract

In this paper we discuss a parallel implementation of an artificial neural network for pattern association, the Bidirectional Associative Memory (BAM). The parallelism inherent in BAM has been exploited using various transputer-based architectures like hypercube, mesh and linear array. The speedup and utilization of the system are obtained for the architectures mentioned above and a comparative study is made. Simulation results show that the hypercube topology gives a better performance in terms of speedup and utilization when compared to other architectures.