The river erosion and deposition algorithm with adaptive search dynamics for balancing exploration and exploitation

Area of Science:

• Artificial Intelligence
• Optimization Algorithms
• Computational Intelligence

Background:

• Optimization is crucial in artificial intelligence (AI), yet balancing exploration and exploitation in complex problems remains challenging.
• Metaheuristic algorithms are commonly used for black-box optimization but often struggle with dynamic equilibrium.
• Existing methods face difficulties in maintaining population diversity and efficiently directing search in intricate solution domains.

Purpose of the Study:

• To introduce the River Erosion and Deposition Algorithm (REDA), a novel search technique for numerical and engineering optimization.
• To address the challenge of balancing exploration and exploitation in complex AI optimization problems.
• To provide a robust algorithm capable of achieving dynamic equilibrium in search processes.

Main Methods:

• Developed the River Erosion and Deposition Algorithm (REDA) with an adaptive search weight for cyclic exploration-exploitation.
• Incorporated a randomized Boolean operator to preserve population diversity.
• Utilized stochastic recombination with an elite memory set for directed search.
• Validated REDA on 19 constrained engineering problems and 29 unconstrained CEC2017 benchmark functions.

Main Results:

• REDA demonstrated significantly superior performance compared to 13 state-of-the-art algorithms.
• Statistical analyses (Friedman, Wilcoxon signed-rank tests) confirmed REDA's high performance, particularly in low-dimensional spaces.
• REDA achieved high accuracy and stability when applied to parameter detection in a solar system model.

Conclusions:

• The proposed River Erosion and Deposition Algorithm (REDA) effectively balances exploration and exploitation in challenging solution domains.
• REDA offers a promising advancement in metaheuristic optimization for AI and engineering applications.
• Experimental validation confirms REDA's robustness, accuracy, and superior performance over existing methods.