Abstract
Many adaptive systems require optimization in real time. Whether it is a robot that must maintain its gait regardless of the terrain or multicore systems needing proper scheduling, optimization is of utmost importance. With hundreds of processes created and evaluated every second, real-time performance optimization is a monumental task. Mother nature has proven that evolution is very effective form of adaptation. Through a stochastic search, i.e. GA, computers harness this power. GAs have been developed to utilize many different parameters, which have a significant effect on the efficiency and effectiveness of a GA. If a GA tasked to optimize these parameters, the result is a rapid and automatic optimization. To test our hypothesis we optimize a GA that solves common optimization functions. The GA's effectiveness is determined by the time it takes to find the solution. Cross validation is utilized, and shows an average 947% performance improvement on training sets and 440% on testing sets. This large improvement in the testing sets shows that an optimized genetic algorithm remains general enough to effectively solve similar problems. (C) 2014 The Authors. Published by Elsevier B.V.