An Open Framework for Constructing Continuous Optimization Problems.

Area of Science:

• Artificial Intelligence
• Optimization Theory
• Computer Science

Background:

• Numerous artificial benchmark problems exist for various continuous optimization types.
• Current methods lack a unified framework, hindering tunable property analysis.
• This limits researchers' ability to thoroughly assess algorithm strengths and weaknesses.

Purpose of the Study:

• To propose a simple and intuitive framework for constructing diverse continuous optimization problems.
• To enable tunable properties for better algorithm analysis.
• To address the limitations of existing, fragmented problem-generation methods.

Main Methods:

• Utilizes a k-d tree to partition the search space.
• Assigns simple functions within each partitioned subspace.
• Implements the framework for global, multimodal, dynamic, and multiobjective optimization scenarios.

Main Results:

• Successfully implemented the framework for various optimization types, including dynamic multiobjective optimization.
• Demonstrated the framework's ability to construct problems with tunable properties.
• Verified the framework's effectiveness using traditional evolutionary algorithms.

Conclusions:

• The proposed framework offers a unified and flexible approach to generating continuous optimization benchmark problems.
• This facilitates more rigorous analysis and comparison of optimization algorithms.
• The framework's adaptability supports research in global, multimodal, dynamic, and multiobjective optimization.