Answers are grounded in this paper and cite their sources.
Evolutionary auxiliary optimization (EAO) methods hold significant promise for practical applications in high-dimensional feature selection. However, current EAO approaches are limited by rigid task generation strategies that rely on single filtering methods, which can restrict performance and increase computational cost. To address these challenges, a novel EAO framework is developed. It employs a multi-indicator integrated task generation strategy, which combines multiple filtering methods to construct complementary auxiliary tasks for eliminating irrelevant features and incorporating complementary knowledge. Additionally, an improved particle swarm optimization algorithm is integrated with a dynamic knowledge transfer mechanism to enhance the optimization in a dual-task scenario. This mechanism incorporates a dynamic transfer factor and a task-aware gene optimization strategy, improving both information exchange efficiency and search capabilities. Empirical studies on 16 publicly available high-dimensional datasets demonstrate the superiority of our proposed EAO over existing state-of-the-art methods. It not only produces superior feature subsets but also significantly reduces runtime.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: