Towards Multi-Objective High-Dimensional Feature Selection via Evolutionary Multitasking

TL;DR

The paper tackles multiobjective feature selection in ultrahigh-dimensional spaces by introducing MO-FSEMT, an Evolutionary Multitasking framework that constructs multiple auxiliary tasks via filtering-based and clustering-based formulations. Each task runs with an independent population and solver, and a task-specific knowledge transfer mechanism exchanges high-quality information across tasks to accelerate convergence and improve solution quality. Empirical results across numerous high-dimensional datasets show MO-FSEMT achieves superior Pareto fronts with fewer selected features and competitive computation times, with ablation studies confirming the value of redundancy removal, diverse formulations, and transfer strategies. This framework advances high-dimensional FS by enabling diverse search landscapes, tailored optimization per task, and explicit, information-rich knowledge transfer, offering practical benefits for scalable feature selection in complex domains.

Abstract

Evolutionary Multitasking (EMT) paradigm, an emerging research topic in evolutionary computation, has been successfully applied in solving high-dimensional feature selection (FS) problems recently. However, existing EMT-based FS methods suffer from several limitations, such as a single mode of multitask generation, conducting the same generic evolutionary search for all tasks, relying on implicit transfer mechanisms through sole solution encodings, and employing single-objective transformation, which result in inadequate knowledge acquisition, exploitation, and transfer. To this end, this paper develops a novel EMT framework for multiobjective high-dimensional feature selection problems, namely MO-FSEMT. In particular, multiple auxiliary tasks are constructed by distinct formulation methods to provide diverse search spaces and information representations and then simultaneously addressed with the original task through a multi-slover-based multitask optimization scheme. Each task has an independent population with task-specific representations and is solved using separate evolutionary solvers with different biases and search preferences. A task-specific knowledge transfer mechanism is designed to leverage the advantage information of each task, enabling the discovery and effective transmission of high-quality solutions during the search process. Comprehensive experimental results demonstrate that our MO-FSEMT framework can achieve overall superior performance compared to the state-of-the-art FS methods on 26 datasets. Moreover, the ablation studies verify the contributions of different components of the proposed MO-FSEMT.

Figures (4)

• Figure 1: Illustration of the proposed MO-FSEMT framework.
• Figure 2: Toy example of the proposed problem formulation strategy.
• Figure 3: Toy example of the proposed task-specific based knowledge transfer mechanism.
• Figure 4: The average computation time of seven compared methods on 26 datasets over 20 independent runs. x-axis: datasets; y-axis: time(s).