Online algorithm selection framework based on the similarity measure

Algorithm selection involves choosing the most appropriate algorithm from a set of candidate algorithms to solve a given problem. However, existing selection methods usually focus on static algorithm selection, ignoring the diversity of problems, and making them less adaptable to unknown problems. In this paper, we propose an online algorithm selection framework (OnlineASF) based on a similarity measure. This framework can dynamically update the algorithm selection model, recommending the best-suited algorithm for unknown problems. We instantiate OnlineASF with six DE algorithms, resulting in the OnlineASF-DE method. Extensive results show that OnlineASF-DE outperforms other methods. The proposed method provides a new approach for online algorithm selection, ensuring higher efficiency in practical applications.