Evolutionary Optimization: A Bird''s Eye View

Many practical decision processes are dependent on the optimization problem solving. Over the last few decades, evolutionary computation has become a popular approach for solving complex optimization problems. Because of their tremendous success, many researchers and practitioners think it has introduced a new field that can be recognized as Evolutionary Optimization (EvOpt). As the optimization (conventional!) is a very old and establish discipline, many concepts have been borrowed from this discipline in the development of evolutionary algorithms and, interestingly, both conventional and evolutionary optimization algorithms are used to solve many similar problems, and their developments have been continuing in parallel. In this talk, we will mainly discuss (i) what optimization really contributes to the real-world decision processes, (ii) what type of optimization problems are well suited for evolutionary algorithms, and (iii) how the evolutionary algorithms are configured for optimization problems solving. For algorithm configuration, we will provide examples of solution representation, initialization, search operators, algorithmic parameters, and algorithm framework. Our experiences with a number of practical issues in defining real-world decision problems and designing their solution approaches will be shared. Note that we will limit our discussions on single-objective optimization but under deterministic, stochastic and time-varying situations.