Challenging the Limits of Binarization: A New Scheme Selection Policy Using Reinforcement Learning Techniques for...

Area of Science:

• Artificial Intelligence
• Computer Science

Background:

• Continuous metaheuristics are challenging to apply to binary problems.
• Binarization schemes are crucial for bridging this gap.
• Novel action selection mechanisms are needed to enhance binarization.

Purpose of the Study:

• Introduce an innovative reinforcement learning policy as an action selection mechanism.
• Apply this policy as a selector for binarization schemes.
• Enhance the application of continuous metaheuristics to binary combinatorial optimization problems.

Main Methods:

• Implemented a novel reinforcement learning policy within a BSS framework.
• Integrated various reinforcement learning and metaheuristic techniques.
• Evaluated the policy on 45 instances of the Set Covering Problem.

Main Results:

• The reinforcement learning policy significantly improved binarization techniques.
• Outperformed traditional methods in precision and efficiency.
• Demonstrated extensibility and adaptability to other techniques and problems.

Conclusions:

• Reinforcement learning is a powerful tool for solving binary combinatorial problems.
• The proposed policy offers significant advantages over traditional methods.
• This approach has broad implications for real-world applications in optimization.