Survey and Comparison of Parallelization Techniques for Genetic Algorithms Cs 294-1 Final Project

This paper surveys a number of parallelization techniques for genetic algorithms, focusing on two: the distributed genetic algorithm and the distributed tness computation. While parallelization of serial algorithms usually requires certain amount of tricks and twitches, the genetic algorithms are parallel by their nature. This feature of genetic algorithms is widely exploited and there exists a number of ways in which a given algorithm may be parallelized. In addition to examining several parallel designs, this paper compares the performance of two implementations (the distributed genetic algorithm and distributed tness computation) on dual-prcessor SunSPARC 20 for optimization of a simple function f (x) = (x c) 10 (described in detail in 6]).