Lion swarm optimization algorithm with transfer learning

Information plays a crucial role in human life, and how to effectively utilize information is an important problem. The Lion Swarm Optimization (LSO) algorithm can be seen as a metaphor for the way lions use information during their hunting behavior. In this algorithm, lions, representing individuals in the swarm, communicate within each cycle, sharing information to approach the optimal solution. However, as LSO performs a global search with particles initialized randomly, it often wastes computational resources and fails to gather useful information from previous searches to help future calculations. Transfer learning, on the other hand, allows information from a source domain to aid in solving problems in the target domain, thus saving computational resources. By integrating transfer learning with LSO, the search process can be accelerated and the search space reduced. In this paper, we propose an LSO algorithm with transfer learning (LSO-TL). We use the k-means clustering algorithm to compute the Maximum Mean Discrepancy (MMD) between the source and target domains to identify the most similar samples, which are then used to initialize the particles of the target domain. The paper also considers dimensional changes during the transfer process, improving the search speed. We test the algorithm using standard benchmark functions to evaluate its effectiveness.