Mapping a class of neural networks on k-ary n-cubes

In this paper we propose a scheme for mapping two important artificial neur al network (ANN) models on the popular k-ary n-cube parallel architectures (KNCs). The scheme is based on generalizing the mapping of a bipartite grap h onto the KNC architecture and thus can be adapted to any model whose comp utations can be represented by a bipartite task graph. Our approach is the first to adjust the granularity of parallelism so as to achieve the best po ssible performance based on properties of the computational model and the t arget architecture. We first introduce a methodology for optimal implementa tion of multi-layer feedforward artificial neural networks (FFANNs) trained with the backpropagation algorithm on KNCs. We prove that our mapping meth odology is time-optimal and that it provides for maximum processor utilizat ion regardless of the structure of the FFANN. We show that the same methodo logy can be utilized for efficient mapping of Radial Basis Function neural networks (RBFs) on KNCs.