{"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/towards-visual-explanations-for-convolutional","title":"Towards Visual Explanations for Convolutional Neural Networks via Input Resampling","arxiv_id":"1707.09641","date":"2017-07-30","proceeding":null,"authors":["Benjamin J. Lengerich","Sandeep Konam","Eric P. Xing","Stephanie Rosenthal","Manuela Veloso"],"abstract":"The predictive power of neural networks often costs model interpretability.\nSeveral techniques have been developed for explaining model outputs in terms of\ninput features; however, it is difficult to translate such interpretations into\nactionable insight. Here, we propose a framework to analyze predictions in\nterms of the model's internal features by inspecting information flow through\nthe network. Given a trained network and a test image, we select neurons by two\nmetrics, both measured over a set of images created by perturbations to the\ninput image: (1) magnitude of the correlation between the neuron activation and\nthe network output and (2) precision of the neuron activation. We show that the\nformer metric selects neurons that exert large influence over the network\noutput while the latter metric selects neurons that activate on generalizable\nfeatures. By comparing the sets of neurons selected by these two metrics, our\nframework suggests a way to investigate the internal attention mechanisms of\nconvolutional neural networks.","url_abs":"http://arxiv.org/abs/1707.09641v2","url_pdf":"http://arxiv.org/pdf/1707.09641v2.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":"towards-visual-explanations-for-convolutional","repo_url":"https://github.com/blengerich/explainable-cnn","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1707.09641","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}