An Effective Knowledge Transfer Approach for Multiobjective Multitasking Optimization.

Area of Science:

• Evolutionary Computation
• Artificial Intelligence
• Optimization Theory

Background:

• Multiobjective multitasking optimization (MTO) is an emerging research area focused on solving related multiobjective problems simultaneously using evolutionary algorithms.
• Knowledge transfer across tasks is crucial for MTO, with positive transfer significantly enhancing performance.
• Existing methods have limited exploration into identifying optimal solutions for effective positive transfer.

Purpose of the Study:

• To propose a novel algorithm for MTO that improves the selection of valuable transferred solutions for positive transfer.
• To enhance the efficiency and effectiveness of knowledge transfer in evolutionary multiobjective multitasking optimization.

Main Methods:

• The proposed algorithm defines positive transfer as a transferred solution being nondominated within its target task.
• It selects neighbors of these positive-transfer solutions as candidates for transfer in subsequent generations.
• The algorithm was evaluated on benchmark MTO problems.

Main Results:

• Numerical studies demonstrated the effectiveness of the proposed approach on benchmark MTO problems.
• The experimental results indicate that the new framework achieves competitive performance compared to existing state-of-the-art MTO methods.
• The strategy of transferring neighbors of nondominated solutions proved beneficial.

Conclusions:

• The proposed algorithm successfully addresses the challenge of finding valuable transferred solutions in MTO.
• This novel approach offers a competitive and effective method for improving positive knowledge transfer in evolutionary multiobjective multitasking optimization.
• The findings suggest a promising direction for future research in MTO.