LaOAS

The fitness landscape of optimization problems significantly impacts the performance of metaheuristic optimization algorithms. While no algorithm performs well on all problems, identifying the most suitable one for a specific problem can be achieved by extracting features from the landscape. However, these features are often extracted before optimization, disregarding valuable knowledge collected during the optimization process. Moreover, existing algorithm selection methods rely on a single algorithm, limiting flexibility and missing potentially better options. To address these challenges, we propose the landscape-aware online algorithm selection (LaOAS) framework, which incorporates a landscape classification technique, an algorithm-switching mechanism, and a dictionary-based context memory method. We instantiate LaOAS with LSHADE-EpSin and LSHADE-cnEpSin, resulting in the LaOAS-LSHADES algorithm, being evaluated across various problems. Extensive experiments using the CEC2013 benchmark and different photovoltaic models show that LaOAS-LSHADES consistently outperforms state-of-the-art algorithms. The LaOAS framework introduces a novel approach to online algorithm selection, improving efficiency in real-world applications.