Inverted Contrastive Learning for unsupervised feature selection

ICLFS treats unsupervised feature selection as a representation learning problem by inverting the data matrix and applying an InfoNCE-based objective to maintain consistency across feature-wise perturbations.