Parallel Processing System for Solution of Simultaneous Linear Equations Based on Discrete-Time Neural Network

Abstract

This paper presents highly parallel multiprocessing system for solution of simultaneous linear equations. The parallel processing system can be naturally discussed based on the concept of discrete-time neural network. Using an algorithm similar to the steepest descent method the system always gives optimal solution of simultaneous linear equations since the state space has no local minima.
The corresponding multiprocessor system is efficiently constructed using ring interconnection network. The practical performance is evaluated by the implementation to demonstrate its advantage.