Asl/nsl: a Multi-level Computational Model for Distributed Neural Simulation*

As neural systems become large and complex, sophisticated tools are needed to support effective model development and efficient simulation processing. Initially, during model development, rich graphical interfaces linked to powerful programming languages and component libraries are the primary requirement. Later, during model simulation, processing efficiency is the primary concern. Workstations and personal computers are quite effective during model development, while parallel and distributed computation become necessary during simulation processing. We give first an overview of modeling and simulation in NSL/ASL together with an example. We then discuss current and future work with the system in the development and simulation of modular neural systems executed in a single computer or distributed computer network. INTRODUCTION In this paper we present an overview of the ASL/NSL computational model for the simulation of neural systems. This work is motivated by the quest to simulate animal-like behavior as faithfully as possible based on existing ethological, physiological and anatomical neural data. At the higher level these models are described in terms of schema (Arbib 1992) modules corresponding to behavior agents (ethology) simulated with ASL (Abstract Schema Language) (Weitzenfeld 1993). At the lower level, neural networks are described in terms of neural modules (physiology and anatomy) simulated with the Neural Simulation Language NSL (Weitzenfeld and Arbib 1994). Many models have been developed and simulated with ASL/NSL (see Weitzenfeld et al. 2000a for examples of models involving perception, visuomotor coordination, motor control as well as technological neural applications).