Sequence-Based Deterministic Initialization for Evolutionary Algorithms.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Heuristic Methods

Background:

• Evolutionary algorithm performance is highly dependent on initial population quality.
• Existing initialization techniques focus on uniform search space coverage but ignore problem-specific information.

Purpose of the Study:

• To propose a novel initialization technique for evolutionary algorithms.
• To develop a heuristic space-filling approach that incorporates function and search space characteristics.
• To improve the performance of computational intelligence algorithms.

Main Methods:

• A new initialization technique integrating function-to-be-optimized and search space properties was developed.
• The technique was tested on unconstrained benchmark problems using various computational intelligence algorithms.
• Performance was evaluated against existing initialization methods.

Main Results:

• The proposed initialization technique significantly improved the performance of all tested algorithms.
• The method provided valuable insights into the behavior of the functions being optimized.
• High-quality initial solutions were generated for certain test problems.
• Benefits were also observed in multiobjective optimization scenarios.

Conclusions:

• The novel initialization technique offers a significant advancement over traditional methods.
• This approach enhances evolutionary algorithm performance and provides functional insights.
• The technique is effective for both single-objective and multiobjective optimization problems.