A novel mating approach for genetic algorithms.

Area of Science:

• Computational Intelligence
• Evolutionary Computation
• Optimization Algorithms

Background:

• Genetic algorithms commonly employ crossover for parent mating, often using random selection.
• Existing mating strategies may lack adaptability to diverse problem landscapes.

Purpose of the Study:

• To introduce a novel parent mating approach for genetic algorithms.
• To develop a flexible framework for diverse mating strategies using a mating index.
• To investigate the performance benefits of the novel approach, particularly the self-adaptive strategy.

Main Methods:

• A new parent mating approach is defined, incorporating a 'mating index' parameter.
• Three mating strategies are proposed: best-first (exploitative), best-last (explorative), and self-adaptive.
• The approach is evaluated using real function optimization, analyzing performance against varying degrees of multimodality.

Main Results:

• Increasing the mating index improves performance as function multimodality increases.
• The self-adaptive mating strategy demonstrates strong performance across multiple case studies.
• The novel approach offers a uniform framework for developing different mating strategies.

Conclusions:

• The proposed novel mating approach enhances genetic algorithm performance, especially in complex optimization tasks.
• The mating index and self-adaptive strategy provide effective control over exploration-exploitation balance.
• This work offers a domain-independent method for improving evolutionary computation.