Comparative evaluation of platforms for parallel Ant Colony Optimization

Guerrero, GD, Cecilia, JM, Llanes, A, Garcia, JM, Amos, M and Ujaldon, M (2014) Comparative evaluation of platforms for parallel Ant Colony Optimization. Journal of Supercomputing, 69. ISSN 0920-8542

Preview

Available under License In Copyright. Download (503kB) | Preview

Abstract

The rapidly growing field of nature-inspired computing concerns the development and application of algorithms and methods based on biological or physical principles. This approach is particularly compelling for practitioners in high-performance computing, as natural algorithms are often inherently parallel in nature (for example, they may be based on a “swarm”-like model that uses a population of agents to optimize a function). Coupled with rising interest in nature-based algorithms is the growth in heterogenous computing; systems that use more than one kind of processor. We are therefore interested in the performance characteristics of nature-inspired algorithms on a number of different platforms. To this end, we present a new OpenCL-based implementation of the Ant Colony Optimization algorithm, and use it as the basis of extensive experimental tests. We benchmark the algorithm against existing implementations, on a wide variety of hardware platforms, and offer extensive analysis. This work provides rigorous foundations for future investigations of Ant Colony Optimization on high-performance platforms.