Parallel Distributed Genetic Programming using Long-term Memory for Dynamic Scheduling Problems

Genetic Programming (GP) is an evolutionary computation method that optimizes the rules defining relationship between environmental states and system output. GP effective for dynamic environments in which information repeatedly changes multiple times. On other hand, GP, are evaluved as state changes, so acquired distant past disappear time, re-learning required a environment. This paper proposes optimization scheduling problem where new jobs arrive intermittently. Specifically, to improve learning efficiency of such periodic environment by dividing population into several subpopulations recording or their characteristics. conducts some numerical experiments on problems irregularly verify usefulness proposed method.