Embedding Q-Learning in the selection of metaheuristic operators: The enhanced binary grey wolf optimizer case
Diego Tapia, Broderick Crawford, Ricardo Soto, Wenceslao Palma, José Lemus-Romani, Felipe Cisternas-Caneo, Mauricio Castillo, Marcelo Becerra, Fernando Paredes, Sanjay Misra · 2021 IEEE International Conference on Automation/XXIV Congress of the Chilean Association of Automatic Control (ICA-ACCA) · 2021
In the different situations present in the industry, combinatorial problems are increasingly frequent. This paper presents the interaction of Metaheuristics and Machine Learning, specifically as Machine Learning can be a support to enhance Metaheuristics. The resolution of the Set Covering Problem is presented, using the Grey Wolf Optimizer and Sine Cosine Algorithm metaheuristics that have been improved by adding a Q-Learning technique for the selection of a Discretization Scheme, using two-steps, intelligently choosing which transfer function to use and which binarization technique to apply in each iteration. The results show a better result for the Grey Wolf Optimizer with Q-Learning configuration, compared to other configurations in the literature, obtaining a better balance between exploration and exploitation.