An Entropy Feedback Based Evolutionary Algorithms and its Application

Evolutionary algorithms (EAs) simulate the process of biological evolution in nature to solve problems. EAs are a class of global optimization methods with high robustness, and have been widely applied in many areas. Feedback control (FC) regulates the behaviors of the system based on feedback. To understand the insight of EAs better and analyze the reported EAs, this paper proposes an entropy feedback based evolutionary algorithms (EFEAs). The entropy is adopted to measure the evolutionary state and acts as the feedback for self-adaptive adjustment. Based on the feedback of entropy, the control parameters, such as ranking selection probability, crossover probability, and mutation probability, are adaptive adjusted to tune the evolutionary state and improve the search performance of EAs. Finally, the state feedback framework is applied to analyze the mechanism and performance of the improved EAs. EFEAs provides a new perspective for understanding the mechanisms of EAs and a new approach to improve the performance of EAs. Experiments in the Travelling Salesman Problem (TSP) show that EFEAs outperforms other comparison algorithms.