Memetic algorithms for continuous optimisation based on local search chains.

Area of Science:

• Computational intelligence
• Optimization algorithms
• Evolutionary computation

Background:

• Memetic algorithms (MAs) are effective for complex continuous optimization problems.
• Intensive continuous local search methods offer high precision but can be computationally expensive.
• Existing MAs may not fully leverage the power of these intensive local search methods.

Purpose of the Study:

• To propose a new memetic algorithm framework that effectively integrates intensive continuous local search methods.
• To introduce the concept of 'local search chain' to enhance the efficiency of local search operators.
• To improve the concentration of local tuning in promising search regions.

Main Methods:

• The proposed memetic algorithm utilizes a 'local search chain' concept, where subsequent local search operations continue from the previous state.
• An instance of the memetic algorithm was implemented using CMA-ES (Covariance Matrix Adaptation Evolution Strategy) as the intensive local search operator.
• Performance was evaluated against other memetic and evolutionary algorithms in the literature.

Main Results:

• The proposed memetic algorithm demonstrated significant benefits in tackling continuous optimization problems.
• Experimental results showed clear superiority, particularly for high-dimensional problems.
• The local search chain mechanism effectively concentrated search efforts in promising areas.

Conclusions:

• The local search chain concept provides an effective way to design memetic algorithms using intensive continuous local search methods.
• The proposed approach offers a computationally efficient and high-precision solution for complex continuous optimization.
• This framework represents a promising advancement for solving high-dimensional optimization challenges.