Implementing radial basis functions neural networks on the systolic MANTRA machine

Abstract

The development of neural network models requires the study o] dedicated hardware architectures. In this paper, we propose an implementation of Radial Basis Function networks, derive an architecture based on an already existing 2D-systolie machine (MANTRA). A systolic algorithm is described to implement the required functions and the suitable sequence of operations. Theoretical effieiencies are estimated on the key tasks and some guidelines ore given for a best usage of the Mantra machine in the studied Framework.