Accelerating The Convergence Of Cubic Interpolation Method Using Concurrent Algorithm Based On The Transputer Technology

Abstract

This paper describes a concurrent method to speed up the convergence of the QIOM (Cubic Interpolation Optimization Method) to obtain the global minima for any nonlinear continuous function. The QIOM is a naturally slow, sequential procedure, but usually guarantees a global solution. The acceleration of convergence is investigated from two points of view. One concurrency can be seen in the region bounded by the lower and upper limits on the function variables. The other concurrency is seen in investigating the QIOM algorithm itself. Due to the fact that many accelerators involve parallel processors, most of the algorithms today used for optimization are written specifically for sequential processors. This algorithm is implemented using Occam and Transputer Technology. The algorithm is tested on one well known nonlinear function.