Towards Autonomous Bound Constraint Handling: Study on an Adaptive Correction in Differential Evolution

This research focuses on the development of adaptive bound constraint handling methods (BCHMs) operator for the Differential Evolution (DE) algorithm in the context of real-valued optimization over bounded feasible domains. The mutation operator involved in DE can generate infeasible solutions that exceed their predefined feasible bounds, necessitating corrective measures. Traditional BCHMs either adjust out-of-bound components individually (component-wise) or revise the entire solution vector (vector-wise) to ensure feasibility. However, the behavior of BCHMs can vary significantly across different fitness landscapes, making it challenging to determine the most suitable approach in a general setting. Drawing from empirical observations on benchmark functions exhibiting non-monotonic infeasibility probability trends, this study argues for the development of adaptive correction operators that can dynamically adjust their behavior based on the problem characteristics.