An improved teaching-learning-based optimization with multi-learning strategy and ranking-based selection for multitask optimization problems

Abstract

In recent years, evolutionary multitask optimization (EMTO) has gained enough attention in the research community. Multitask optimization (MTO) makes full use of the potential parallelism of population search-based efforts to achieve cross-domain optimization of multiple optimization problems and make knowledge migration between different optimization problems possible. Whether from the angle of convergence speed or quality, it shows better ability than single task optimization. However, we note that the current EMTO algorithm uses only a single crossover for knowledge transfer, which can’t reduce the impact of negative information in the process of knowledge transfer. There are some limitations in solving available tasks. This paper presents a new multitask teaching-learning-based optimization with multi-learning strategies and ranking-based selection strategy (MTTLBO-MR). In the teaching stage, there are three learning strategies for students to choose. In the learning stage, each student chooses the learning strategy from two strategies according to their own situation. Different from the greedy selection strategy of the original TLBO, the new offspring is selected on the basis of the rank based on the factorial rank and diversity rank of each individual. The test result states clearly that MTTLBO-MR updates knowledge through suitable learning strategies at different optimization stages and has outstanding performance on dealing with multitask optimization tasks.