A Machine-Independent Analysis of Parallel Genetic Algorithms
V. Scott Gordon
Computer and Information Science,
Sonoma State University,
Rohnert Park, CA 94928, USA
Darrell Whitley
Department of Computer Science,
Colorado State University,
Fort Collins, CO 80523, USA
Abstract
This paper presents a machine-independent study of parallel genetic algorithm performance. Our approach utilizes a dataflow model of computation in conjunction with Sisal, an implicit parallel programming language. We compare problem-solving power and runtime efficiency for several parallel genetic algorithms under uniform conditions. The proposed method makes it possible to identify all sources of potential parallelism, and to locate and measure bottlenecks. The dataflow model thus provides a systematic way to develop and evaluate genetic algorithms.
