New Evolutionary Algorithm Framework OKAEM Unifies Knowledge Transfer and Adaptation

A novel evolutionary framework called the Optimization Knowledge Adaptation Evolutionary Model (OKAEM) has been introduced to address limitations in current evolutionary algorithm research. Unlike existing approaches that treat knowledge transfer and online adaptation separately, OKAEM provides a unified learnable framework capable of both capabilities simultaneously. The model parameterizes evolutionary operators using attention mechanisms, enabling adaptive updates based on available optimization knowledge. This approach aims to overcome the incomplete utilization of prior knowledge seen in evolutionary sequential transfer optimization, while also moving beyond adaptive strategies that are limited to tailoring specific operators. By encapsulating optimization knowledge from historical populations and fitness evaluations during iterative search processes, OKAEM facilitates more effective knowledge transfer and adaptation. The framework represents arXiv preprint 2501.02200v2, which was announced as a replacement cross submission.