An Estimation of Distribution Algorithm for Large-Scale Optimization with Cooperative Co-evolution and Local Search

Abstract

Cooperative co-evolution (CC) is an effective framework for evolutionary algorithms (EAs) to solve large-scale optimization problems. By combining a divide-and-conquer strategy and the classic evolutionary algorithms (EA) like genetic algorithm (GA), CC has shown promising performance in many fields. As a family of EAs, the estimation of distribution algorithm (EDA) is good at search diversity maintenance, but its capability in solving large-scale problems has not been fully explored. In this paper, we aim to propose a new estimation of distribution algorithm with the cooperative co-evolution framework (EDACC). The proposed EDACC has the following features. 1) The differential grouping (DG) strategy is applied for variable decomposition. 2) A combination of the Gaussian and Cauchy distributions are adopted to generate offspring. 3) A local search method is performed in promising domains to accelerate the search. To verify the performance of EDACC, experiments are conducted on 20 single-objective functions in the CEC 2010 benchmarks. The experimental results show that EDACC can still achieve competitive performance in spite of the weakness of the original EDAs like the low accuracy in global optima searching compared with classical EAs.

Publication
International Conference on Neural Information Processing
Date

The slide is not available in public now. Drop me an email if you’re interested.

More information can be found in this link .