A quasi-parallel realization of the investment frontier in computer resource allocation using simple genetic algorithm on a single computer

Abstract

The concept of portfolio management of algorithm is implemented in a new architecture based on ideas of cooperating multi-agents. Each agent is a simple genetic algorithm with identical structure but possibly different parameters. We introduce a "resource allocation vector" to coordinate the computing resources allocated to each agent. We also encourage constructive collaboration among agents by the exchange of the individuals in the population of each genetic algorithm using an individual-migration matrix. The algorithm can be implemented in a serial computer and behaves statistically in a quasi-parallel manner. We have performed extensive statistical analysis using measures such as the mean and variance of the first passage time to solution. The existence of investment frontier in solving the Schaffer function problem is demonstrated and application to the solving of the traveling salesman problem is discussed. The results suggest a more effective way to utilize computing resources.