Various Degrees of Steadiness in NSGA-II and Their Influence on the Quality of Results
Maxim Buzdalov, Vladimir Parfenov · 2015
Steady-state evolutionary algorithms are often favoured over generational ones due to better scalability in parallel and distributed environments. However, in certain conditions they are able to produce results of better quality as well. We consider several ways to introduce various ``degrees of steadiness'' in the NSGA-II algorithm, some of which have not been known in literature, and show experimentally (on a corpus of 21 test problems) the presence of a general trend: algorithms with more steadiness yield better results.