A Q-Learning-Based Brainstorming Optimization Algorithm for Solving Multimodal Optimization Problems

Large models involve solving more complex problems using intelligent computing methods, such as machine learning, evolutionary algorithms, and swarm intelligence. Multiple choices are a common phenomenon in consumer electronics. Multimodal optimization (MMO) aims to find multiple satisfied solutions simultaneously. The number (quantity) and the accuracy (quality) of solutions are equally important in solving multimodal optimization problems (MMOP). This paper proposes a Q-learning-based brainstorming optimization algorithm with two niching (QLBSO2N) strategies to solve MMOPs. The proposed algorithm has three main components: two niching strategies, Q-learning aided solutions generation, and evolution. Two niching strategies, i.e., crowding and speciation, which replace the k-means clustering strategy, divide the population into several subpopulations. The Q values from the Q-learning method, instead of a random value in the original BSO algorithm, have been used to determine the new solution generation. With the Q-learning method, each individual is regarded as an agent that adaptively selects appropriate actions in the current state. The learning-aided solution evolution (LSE) strategy, which has a robust global search ability, evolves new individuals and enhances population convergence. Thirty nonlinear equation systems are selected to validate the performance of the QLBSO2N algorithm by comparing it with the other four state-of-the-art algorithms. The experimental results show that QLBSO2N outperforms the compared algorithms on most test functions.