A Fireworks Algorithm Based on Transfer Spark for Evolutionary Multitasking

Area of Science:

• Computational Intelligence
• Evolutionary Computation
• Optimization Algorithms

Background:

• Multitask optimization problems require algorithms to solve multiple tasks simultaneously.
• Evolutionary algorithms improve efficiency through genetic complementarity between tasks.
• Existing methods face challenges in effectively transferring information across tasks.

Purpose of the Study:

• To propose a novel multitask fireworks algorithm (MFA) for solving multitask optimization problems.
• To introduce innovative transfer sparks for effective genetic information exchange.
• To enhance the performance of evolutionary multitasking algorithms.

Main Methods:

• Development of a novel multitask fireworks algorithm (MFA).
• Generation of transfer sparks with adaptive length and direction vectors for inter-task information transfer.
• Comparative analysis against state-of-the-art evolutionary multitasking algorithms.

Main Results:

• The proposed MFA demonstrates superior performance compared to existing algorithms.
• Effective transfer of genetic information via novel transfer sparks was observed.
• Improved results on both single-objective and multiobjective MTO test suites.

Conclusions:

• The novel multitask fireworks algorithm offers a significant advancement in evolutionary multitasking.
• Transfer sparks are crucial for enhancing information exchange and algorithm performance.
• The proposed MFA provides a promising solution for complex optimization challenges.