Computational Evidence on Genetic Algorithms and Reinforcement Learning Algorithms for Module

Abstract

In a distributed system, it is important to find an assignment of program modules to processors such that system cost is minimized or system throughput is maximized. Researchers have proposed serveral versions of formulation to this problem. Howerver, most of the versions proposed are NP-complete, and thus finding the exact solutions is computationally intractable. In this paper, we propose a genetic algorithm and a reinforcement learning algorithm to find the near-optimal module assignment. We present the computational evidence of the two algorithms with a set of simulated data. The direction of furture research is suggested according to the experimental results.