Experience with the development of a neural network simulator

Abstract

Researchers experimenting with structured neural networks must be able to imple­ment and test their ideas. Since most of the network models presume much structure, and network architecture design is a major component in the research effort, it is necessary to develop a general-purpose parallel neural network simulation environ­ment that supports a variety of architectures. In this article the development of a general-purpose parallel neural network simulator called SIMNET is described. It is developed to support Parallel Distributed Processing research. Its major goal is to fa­cilitate simulation by providing both the tools for network description and a powerful user interface for design and analysis of experiments.