Adaptively weighted decomposition based multi-objective evolutionary algorithm

Multi-objective evolutionary algorithm based on Decomposition (MOEA/D) decomposes a multi-objective problem into a number of scalar optimization problems using uniformly distributed weight vectors. However, uniformly distributed weight vectors do not guarantee uniformity of solutions on approximated Pareto-Front. This study proposes an adaptive strategy to modify these scalarizing weights after regular intervals by assessing the crowdedness of solutions using crowding distance operator. Experiments carried out over several benchmark problems with complex Pareto-Fronts show that such a strategy helps in improving the convergence and diversity of solutions on approximated Pareto-Front. Proposed algorithm also shows better performance when compared with other state-of-the-art multi-objective algorithms over most of the benchmark problems.

Automated summary

This study proposes an adaptive strategy to modify scalarizing weights in multi-objective evolutionary algorithm based on Decomposition (MOEA/D) after regular intervals by assessing the crowdedness of solutions using crowding distance operator. The experiments show that this strategy improves the convergence and diversity of solutions on approximated Pareto-Front, and the proposed algorithm performs better than other state-of-the-art multi-objective algorithms on most benchmark problems.