{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/maximally-invariant-data-perturbation-as","title":"Maximally Invariant Data Perturbation as Explanation","arxiv_id":"1806.07004","date":"2018-06-19","proceeding":null,"authors":["Satoshi Hara","Kouichi Ikeno","Tasuku Soma","Takanori Maehara"],"abstract":"While several feature scoring methods are proposed to explain the output of\ncomplex machine learning models, most of them lack formal mathematical\ndefinitions. In this study, we propose a novel definition of the feature score\nusing the maximally invariant data perturbation, which is inspired from the\nidea of adversarial example. In adversarial example, one seeks the smallest\ndata perturbation that changes the model's output. In our proposed approach, we\nconsider the opposite: we seek the maximally invariant data perturbation that\ndoes not change the model's output. In this way, we can identify important\ninput features as the ones with small allowable data perturbations. To find the\nmaximally invariant data perturbation, we formulate the problem as linear\nprogramming. The experiment on the image classification with VGG16 shows that\nthe proposed method could identify relevant parts of the images effectively.","url_abs":"http://arxiv.org/abs/1806.07004v2","url_pdf":"http://arxiv.org/pdf/1806.07004v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"maximally-invariant-data-perturbation-as","repo_url":"https://github.com/sato9hara/PertMap","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1806.07004","atlas_url":"https://app.syntology.ai/?focus=1806.07004","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}