WCSE

Deep Evolutionary Algorithm for Large-Scale Sequence Optimization

Abstract— Aiming at the problem of slow convergence and poor effect of GA in solving large-scale sequence optimization problems, a deep evolutionary algorithm is proposed in this work. The algorithm uses the trained network to quickly find the initial solution of the problem and injects the initial solution into the GA population for further optimization. Finally, the optimal solution in the final population will be further optimized by the 2-OPT. The proposed algorithm is compared with other algorithms on the tsplib95 instances and industrial sorting sequence optimization dataset. Experimental results show that the proposed algorithm achieves the best optimization performance compared with other algorithms, especially for large-scale sequence optimization problems.

Index Terms— genetic algorithm; deep reinforcement learning; 2-OPT; large-scale sequence optimization.

Detian Zeng
Information Institute, Hunan University of Humanities, Science and Technology, China; School of Computer, National University of Defense Technology, China
Gaofeng Zhu
Information Institute, Hunan University of Humanities, Science and Technology, China
Qiucheng Miao
School of Computer, National University of Defense Technology, China
Hao Liu
Information Institute, Hunan University of Humanities, Science and Technology, China

[Download]

Cite: Detian Zeng, Gaofeng Zhu, Qiucheng Miao, Hao Liu, " Deep Evolutionary Algorithm for Large-Scale Sequence Optimization, " WCSE 2022 Spring Event: 2022 9th International Conference on Industrial Engineering and Applications, pp. 1493-1500, Sanya, China, April 15-18, 2022.