基于分解的多目标进化算法分析.docx

Abstract
The optimization problems in real life tend to cope with multiple targets, for example, time, cost, quality, quantity, etc. Evolutionary algorithms are a class of stochastic search algorithm, by simulating the natural biological evolution and natural selection to guide an evolutionary process of individuals and to accumulate the dominant genes. Evolutionary algorithms are very robust and strong adaptive, which can help us to avoid much limitations when solving problems directly only with some classic mathematic approaches, thus evolutionary algorithms can deal with plex problems more effectively than traditional optimization methods feel difficult to solve. What’s more, big esses of NSGA-II[1], an excellent evolutionary algorithm based on the Pareto dominance method, give researchers a false impression that traditional mathematic methods in optimization domain seem out of date. Fortunately, in recent years, methods based on scalar functions and position approaches begin to change this situation and e one of most novel as well as attractive trends originating from traditional mathematic theory, then affects further to the entire multi-objective optimization domain. As result, to find inspirations among traditional principles and approaches has e a new path. Usually, the sub-objectives among one multi-objective optimization problem are conflictive between each other. In the meanwhile, optimization algorithm is aiming to find a set promised solutions to the each sub-problem, ., the solutions named Pareto optima. Supposed to be given an aggregative function and some orders with several preferences, a multi-objective problem can be converted to a single-objective optimization problem that means many essful single-objective optimization algorithms can be easily adopted to explore optimal solutions. However, an ideal trade-off relation for each sub-objective cannot be predetermined, therefore an approach can provide the best candidate solutions and choose pref