{"ID":2851229,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2510.20627","arxiv_id":"2510.20627","title":"H-SPLID: HSIC-based Saliency Preserving Latent Information Decomposition","abstract":"We introduce H-SPLID, a novel algorithm for learning salient feature representations through the explicit decomposition of salient and non-salient features into separate spaces. We show that H-SPLID promotes learning low-dimensional, task-relevant features. We prove that the expected prediction deviation under input perturbations is upper-bounded by the dimension of the salient subspace and the Hilbert-Schmidt Independence Criterion (HSIC) between inputs and representations. This establishes a link between robustness and latent representation compression in terms of the dimensionality and information preserved. Empirical evaluations on image classification tasks show that models trained with H-SPLID primarily rely on salient input components, as indicated by reduced sensitivity to perturbations affecting non-salient features, such as image backgrounds. Our code is available at https://github.com/neu-spiral/H-SPLID.","short_abstract":"We introduce H-SPLID, a novel algorithm for learning salient feature representations through the explicit decomposition of salient and non-salient features into separate spaces. We show that H-SPLID promotes learning low-dimensional, task-relevant features. We prove that the expected prediction deviation under input pe...","url_abs":"https://arxiv.org/abs/2510.20627","url_pdf":"https://arxiv.org/pdf/2510.20627v2","authors":"[\"Lukas Miklautz\",\"Chengzhi Shi\",\"Andrii Shkabrii\",\"Theodoros Thirimachos Davarakis\",\"Prudence Lam\",\"Claudia Plant\",\"Jennifer Dy\",\"Stratis Ioannidis\"]","published":"2025-10-23T15:02:07Z","proceeding":"cs.LG","tasks":"[\"cs.LG\"]","methods":"[]","has_code":false,"code_links":[{"ID":607880,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_id":2851229,"paper_url":"https://arxiv.org/abs/2510.20627","paper_title":"H-SPLID: HSIC-based Saliency Preserving Latent Information Decomposition","repo_url":"https://github.com/neu-spiral/H-SPLID","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}