{"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/on-the-robustness-of-convolutional-neural","title":"On the Robustness of Convolutional Neural Networks to Internal Architecture and Weight Perturbations","arxiv_id":"1703.08245","date":"2017-03-23","proceeding":null,"authors":["Nicholas Cheney","Martin Schrimpf","Gabriel Kreiman"],"abstract":"Deep convolutional neural networks are generally regarded as robust function\napproximators. So far, this intuition is based on perturbations to external\nstimuli such as the images to be classified. Here we explore the robustness of\nconvolutional neural networks to perturbations to the internal weights and\narchitecture of the network itself. We show that convolutional networks are\nsurprisingly robust to a number of internal perturbations in the higher\nconvolutional layers but the bottom convolutional layers are much more fragile.\nFor instance, Alexnet shows less than a 30% decrease in classification\nperformance when randomly removing over 70% of weight connections in the top\nconvolutional or dense layers but performance is almost at chance with the same\nperturbation in the first convolutional layer. Finally, we suggest further\ninvestigations which could continue to inform the robustness of convolutional\nnetworks to internal perturbations.","url_abs":"http://arxiv.org/abs/1703.08245v1","url_pdf":"http://arxiv.org/pdf/1703.08245v1.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":"on-the-robustness-of-convolutional-neural","repo_url":"https://github.com/kreimanlab/neural_net_robustness","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[],"methods":[{"method_slug":"1x1-convolution","method_name":"1x1 Convolution"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"grouped-convolution","method_name":"Grouped Convolution"},{"method_slug":"local-response-normalization","method_name":"Local Response Normalization"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1703.08245","atlas_url":"https://app.syntology.ai/?focus=1703.08245","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}