OPTIMIZATION METHOD AND SYSTEM BASED ON EXPENSIVE MULTI-OBJECTIVE OPTIMIZATION PROBLEMS

摘要

An optimization method and system based on expensive multi-objective optimization problems. The method comprises: if a plurality of objectives in a task simultaneously meets a preset requirement, transforming objective functions corresponding to a plurality of irrelevant objectives by mapping (S100); transforming the objective functions into a group of task series with relevance or similarity by mapping to associate the plurality of irrelevant objective functions (S200); and carrying out joint modeling on the task series with relevance or similarity to generate a multi-task Gaussian process model. According to the method, in an optimization link, an established multi-task Gaussian process model is employed to predict each ask, then a task prediction value is inversely mapped to an objective prediction value to achieve the purpose of avoiding the calculation of an expensive objective function, and finally, the optimization result is output. The optimization cost of the plurality of irrelevant objectives is lowered.