{"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/influence-directed-explanations-for-deep","title":"Influence-Directed Explanations for Deep Convolutional Networks","arxiv_id":"1802.03788","date":"2018-02-11","proceeding":"ICLR 2018 1","authors":["Klas Leino","Shayak Sen","Anupam Datta","Matt Fredrikson","Linyi Li"],"abstract":"We study the problem of explaining a rich class of behavioral properties of\ndeep neural networks. Distinctively, our influence-directed explanations\napproach this problem by peering inside the network to identify neurons with\nhigh influence on a quantity and distribution of interest, using an\naxiomatically-justified influence measure, and then providing an interpretation\nfor the concepts these neurons represent. We evaluate our approach by\ndemonstrating a number of its unique capabilities on convolutional neural\nnetworks trained on ImageNet. Our evaluation demonstrates that\ninfluence-directed explanations (1) identify influential concepts that\ngeneralize across instances, (2) can be used to extract the \"essence\" of what\nthe network learned about a class, and (3) isolate individual features the\nnetwork uses to make decisions and distinguish related classes.","url_abs":"http://arxiv.org/abs/1802.03788v2","url_pdf":"http://arxiv.org/pdf/1802.03788v2.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":"influence-directed-explanations-for-deep","repo_url":"https://github.com/cmu-transparency/lib-attribution","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"influence-directed-explanations-for-deep","repo_url":"https://github.com/pytorch/captum","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"BSD-3-Clause"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1802.03788","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}