A multitask optimization algorithm based on elite individual transfer.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Machine Learning

Background:

• Evolutionary multitasking algorithms solve multiple optimization tasks simultaneously, leveraging knowledge transfer for efficiency.
• Current methods face challenges in optimizing knowledge transfer, leading to potential negative transfer between tasks.

Purpose of the Study:

• To propose a novel single-objective multitask optimization algorithm, MSOET, enhancing knowledge transfer.
• To improve positive transfer and mitigate negative transfer between optimization tasks.

Main Methods:

• Developed MSOET, a multitask optimization algorithm based on elite individual transfer.
• Implemented a probabilistic approach for knowledge transfer execution.
• Utilized current populations and elite individuals to construct a Gaussian distribution model for offspring generation.

Main Results:

• MSOET demonstrated superior performance compared to ten other multitask optimization algorithms.
• Experimental results confirmed the algorithm's excellent performance and strong robustness across tasks.

Conclusions:

• The proposed MSOET effectively improves knowledge transfer in evolutionary multitasking.
• The algorithm shows significant potential for enhancing efficiency and robustness in solving multiple optimization problems simultaneously.