Mutation strategy improves GAs performance on epistatic problems

We examine the behavior of a parallel varying mutation genetic algorithm (GA) on epistatic problems using NK-landscapes. We discuss properties of NK-landscapes and show that mutation strategy is an important factor to improve the performance of GAs on epistatic problems. The effect of (extinctive) selection is also highlighted. Similar to recent works, we conduct our study on relatively larger landscapes than previous studies in order to be a step closer to problems found in real world applications