Random generator quality and GP performance
Mark M. Meysenburg, James A. Foster · 1999
In previous studies, the authors found that pseudo-random number generator (PRNG) quality had little effect on the performance of a simple genetic algorithm (GA). This paper extends our work to the area of genetic programming (GP). We examine the effect of PRNG quality on the performance of GP techniques. We detail a set of PRNGs which generate random numbers through various techniques, and a method for evaluating the quality of these PRNGs. We explain the application of detailed statistical analysis to the results of many individual GP runs, over a set of four GP test problems. We found no evidence to support the notion that higher quality PRNGs caused improved GP performance.