CLMOAS:Collaborative  Large-scale Multi-objective Optimization Algorithms with Adaptive Strategies

Read the full article See related articles

Listed in

This article is not in any list yet, why not save it to one of your lists.
Log in to save this article

Abstract

In the field of multi-objective evolutionary optimization, existing research has mostly focused on the scalability of the objective dimension, while insufficient attention has been paid to the scalability of the decision variable dimension. However, in many practical application scenarios, complex optimization problems with the co-existence of multi-objectives and large-scale decision variables are often faced. In consideration of this, this paper comes up with a novel large-scale multi-objective evolutionary optimization algorithm, the core idea of which is to classify the decision variables by clustering method, and on the basis of which the LMEA algorithm is improved, a new dominance relation is introduced, and the CLMOAS algorithm is constructed, aiming at effectively solving the dominance-resistance problem in the traditional dominance relation. The algorithm first utilizes the clustering technique to classify the decision variables into two categories: those related to convergence and those related to diversity. For these two categories of variables,Various optimization approaches have been developed to achieve targeted optimization. In the diversity-related strategy, a novel angle-based dominance relationship is introduced to reduce the dominance resistance encountered by the algorithm during the optimization process, so as to enhance the optimization efficiency and performance performance of the algorithm. To verify the performance advantages of the proposed algorithm, the paper performs comparative experiments between the LMEA algorithm and several other representative multi - objective evolutionary algorithms across multiple mainstream multi - objective optimization test sets. The experimental results show that the CLMOAS algorithm shows better performance than the original evolutionary algorithms on most of the test sets, verifying its effectiveness and superiority in solving multi-objective and large-scale decision variable problems.

Article activity feed