Fast and Efficient Black Box Optimization Using the Parameter-less Population Pyramid.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Evolutionary Computation

Background:

• Traditional evolutionary optimization methods often require extensive parameter tuning.
• Existing parameter-less methods may compromise performance for applicability.
• A need exists for effective optimization techniques that are general and easy to use.

Purpose of the Study:

• To introduce and evaluate the parameter-less population pyramid (P3) method for black-box optimization.
• To demonstrate P3's ability to optimize complex problems without user-specified parameters.
• To compare P3's performance against advanced state-of-the-art algorithms.

Main Methods:

• Developed the parameter-less population pyramid (P3) approach, replacing generational models with iterative population creation and expansion.
• Integrated local search and advanced crossover techniques within the P3 framework.
• Tested P3 across seven diverse problems and an average of 18 problem sizes each.

Main Results:

• P3 significantly outperformed five advanced comparison algorithms in terms of evaluations needed to find the global optimum and final fitness.
• The method demonstrated effective maintenance, addition, and exploitation of diversity.
• P3 achieved high performance without problem-specific tuning, unlike other leading algorithms.

Conclusions:

• P3 is an efficient and general parameter-less approach for black-box optimization.
• The method offers superior effectiveness compared to existing state-of-the-art techniques.
• P3 successfully balances performance quality with broad applicability.