HALA: A Hybrid Dual-Population Optimizer Integrating an Enhanced Artificial Lemming Algorithm and SHADE

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Metaheuristics

Background:

• Intelligent systems face complex optimization challenges.
• Existing metaheuristics like Artificial Lemming Algorithm (ALA) struggle with premature convergence and local refinement.
• High-dimensional multimodal environments pose significant difficulties for current algorithms.

Purpose of the Study:

• Introduce HALA, a novel hybrid dual-subpopulation optimizer.
• Address limitations of ALA and other metaheuristics in complex optimization scenarios.
• Enhance exploration and exploitation balance for improved performance.

Main Methods:

• Integrates an enhanced Artificial Lemming Algorithm (ALA) with the Success-History Based Adaptive Differential Evolution (SHADE) algorithm.
• Employs two interacting subpopulations: one for exploration (enhanced ALA with t-distribution and Levy flight), another for exploitation (SHADE with adaptive archive).
• Utilizes periodic bidirectional elite migration for knowledge transfer between subpopulations.

Main Results:

• HALA demonstrates competitive or superior solution quality and faster convergence compared to 17 advanced metaheuristics on IEEE CEC2017 and CEC2022 benchmark suites.
• Achieves favorable Friedman average rankings across various dimensions (10, 30, 50, 100) and test suites.
• Successfully solves five constrained engineering design problems, yielding optimal or near-optimal results.

Conclusions:

• HALA effectively overcomes limitations of individual algorithms like ALA and SHADE.
• The hybrid approach provides robust stability and superior performance in challenging optimization tasks.
• HALA shows significant potential for practical applications in engineering optimization.