Genetic Algorithms

Abstract

Optimization algorithms for blackbox functions can be broadly split into two categories: heuristic and non-heuristic. A heuristic is any approach that, while supported by some argument of why it should succeed, does not include a guarantee of success. In the framework of blackbox optimization, we take this statement to mean that the algorithm does not employ mathematically proven stopping criterion. Hence, derivative-free optimization is distinguished from heuristic blackbox optimization, by the presence of a stopping test that provides some assurance of optimality.