Improved Snake Optimizer Using Sobol Sequential Nonlinear Factors and Different Learning Strategies and Its Applications

• Published in: Mathematics (Basel) 2024-06, Vol.12 (11), p.1708
• Main Authors: Zheng, Wenda, Ai, Yibo, Zhang, Weidong
• Format: Article
• Language: English
• Subjects: Algorithms; Analysis; Animal reproduction; Design engineering; Discretization; Effectiveness; Evolutionary computation; Exploitation; feature selection; Females; Food; Genetic algorithms; Heuristic methods; Heuristic programming; learning strategy; Machine learning; Males; Mathematical models; Mathematical optimization; Methods; nonlinear factor; Nonlinear theories; Optimization; Optimization techniques; Performance evaluation; real-world optimization; Sequences; Sequential analysis; snake optimizer; Snakes; Sobol sequences
• ISSN: 2227-7390; 2227-7390
• DOI: 10.3390/math12111708

The Snake Optimizer (SO) is an advanced metaheuristic algorithm for solving complicated real-world optimization problems. However, despite its advantages, the SO faces certain challenges, such as susceptibility to local optima and suboptimal convergence performance in cases involving discretized, high-dimensional, and multi-constraint problems. To address these problems, this paper presents an improved version of the SO, known as the Snake Optimizer using Sobol sequential nonlinear factors and different learning strategies (SNDSO). Firstly, using Sobol sequences to generate better distributed initial populations helps to locate the global optimum solution faster. Secondly, the use of nonlinear factors based on the inverse tangent function to control the exploration and exploitation phases effectively improves the exploitation capability of the algorithm. Finally, introducing learning strategies improves the population diversity and reduces the probability of the algorithm falling into the local optimum trap. The effectiveness of the proposed SNDSO in solving discretized, high-dimensional, and multi-constraint problems is validated through a series of experiments. The performance of the SNDSO in tackling high-dimensional numerical optimization problems is first confirmed by using the Congress on Evolutionary Computation (CEC) 2015 and CEC2017 test sets. Then, twelve feature selection problems are used to evaluate the effectiveness of the SNDSO in discretized scenarios. Finally, five real-world technical multi-constraint optimization problems are employed to evaluate the performance of the SNDSO in high-dimensional and multi-constraint domains. The experiments show that the SNDSO effectively overcomes the challenges of discretization, high dimensionality, and multi-constraint problems and outperforms superior algorithms.