Adaptive collaborative correlation learning-based semi-supervised multi-label feature selection

doi:10.1016/j.patcog.2026.114176

Adaptive collaborative correlation learning-based semi-supervised multi-label feature selection

2026-09-29

0

PRE

AI

DOI：10.1016/j.patcog.2026.114176

Original

Original request for help

Save

Abstract

En 中文

Semi-supervised multi-label feature selection has recently been developed to solve the curse of dimensionality problem in high-dimensional multi-label data with certain samples missing labels. Although many efforts have been made, most existing methods use a predefined graph approach to capture the sample similarity or the label correlations. In this manner, the presence of noise and outliers within the original feature space can undermine the reliability of the resulting sample similarity graph. In addition, the predefined graph approach also fails to precisely depict label correlations due to the presence of unknown labels. Moreover, these methods primarily focus on the discriminative power of selected features, while neglecting their redundancy. In this paper, we propose an Adaptive Collaborative Correlation lEarning-based Semi-Supervised Multi-label Feature Selection (Access-MFS) method to address these issues. Specifically, we develop a generalized regression model equipped with an extended uncorrelatedness constraint to select discriminative features with reduced redundancy, while maintaining consistency between the predicted and ground-truth labels for labeled instances. Within this framework, instance similarity and label correlation are adaptively learned, enabling the sample similarity graph and the label similarity graph to be collaboratively refined throughout the optimization process. The jointly learned graphs preserve local sample relationships, encouraging similar samples to have similar labels, while promoting consistent predictions among strongly correlated labels. An efficient algorithm is designed to solve the proposed optimization model, along with analysis of its convergence and complexity. Extensive experimental results demonstrate the superiority of the proposed Access-MFS over other state-of-the-art methods.

Keywords:

Feature selection

Semi-supervised multi-label learning

Generalized regression model

Adaptive similarity graph learning

Adaptive collaborative correlation learning-based semi-supervised multi-label feature selection

Abstract

Journal

Organization

Cited Papers