Multiobjective Multitask Evolutionary Optimization based on M2M Decomposition with Dynamic Resource Allocation

Multiobjective multitask optimization (MO-MTO) algorithms aim to solve a set of multiobjective optimization problems (MOPs) simultaneously in a single run. This paper proposed an innovative MO-MTO algorithm based on M2M decomposition and dynamic resource allocation (MMTOID-DRA) for the first time. In this algorithm, each task in an MO-MTO problem is decomposed into a series of constrained multiobjective subproblems to ensure population diversity. Furthermore, during the iteration process, more computational resources are dynamically allocated to promising subregions to accelerate convergence. MMTO/D-DRA was comprehensively tested on nine benchmark problems. The experimental results compared it with five state-of-the-art MO-MTO algorithms, demonstrating the superiority of the proposed algorithm.