Paper Details

Nonconvex continuous optimization experiments on a transputer system

Authors: Laak, A. ter, Hertzberger, L. O., Sloot, P. M. A.

Abstract:

In this paper we investigate the functionality of various parallel implementations of Simulated Annealing on a transputer platform. The optimization problem to be solved is that of efficiently finding the global minimum in continuous spaces. Our work concentrates on the consequences of long-range and short-range interactions on algorithmic and geometric decomposition schemes. We introduce a mixed transputer topology to by-pass some of the inherent time critical operations involved. We show that combining the Fast Simulated Annealing algorithm with a systolic decomposition strategy results in a highly efficient algorithm for continuous optimization problems. Experiments indicate that incorporation of functional decomposition of the energy function results in a near optimal implementation.

Proceedings:

Proceedings of WoTUG-15: Transputer Systems - ongoing Research, Alastair R. Allen, 1992, pp 251 - 265 published by IOS Press, Amsterdam

This record in other formats: