Exact solutions for evolutionary strategies on harmonic landscapes

Area of Science:

• Computational intelligence
• Evolutionary computation

Background:

• Evolutionary strategies (ES) are optimization algorithms inspired by natural evolution.
• Understanding the impact of selection mechanisms on ES performance is crucial for algorithm design.

Purpose of the Study:

• To compare the performance of two distinct evolutionary strategies: a thermodynamic (local selection) and a biological (global selection) approach.
• To analyze their behavior and derive analytical solutions within harmonic landscapes.

Main Methods:

• Implementation of two evolutionary strategies based on ensembles of searchers.
• Utilizing harmonic landscapes as test potentials for evaluating strategy performance.
• Derivation of analytical solutions for specific initial conditions.

Main Results:

• Both thermodynamic and biological strategies yield analytical solutions of the same type for harmonic potentials.
• Exact comparison of the velocities of the two strategies is possible through these solutions.
• A mixed strategy combining local and global selection is also explored.

Conclusions:

• The study provides a theoretical framework for comparing different selection mechanisms in evolutionary computation.
• Analytical solutions offer precise insights into the dynamics of evolutionary search.
• The findings contribute to the development of more efficient optimization algorithms.