DC Field | Value | Language |
---|---|---|
dc.contributor.author | Jeong, Jongheon | ko |
dc.contributor.author | Shin, Jinwoo | ko |
dc.date.accessioned | 2020-12-11T07:10:18Z | - |
dc.date.available | 2020-12-11T07:10:18Z | - |
dc.date.created | 2020-12-02 | - |
dc.date.created | 2020-12-02 | - |
dc.date.issued | 2020-12-07 | - |
dc.identifier.citation | 34th Conference on Neural Information Processing Systems (NeurIPS) 2020 | - |
dc.identifier.uri | http://hdl.handle.net/10203/278234 | - |
dc.description.abstract | A recent technique of randomized smoothing has shown that the worst-case (adversarial) ℓ 2 -robustness can be transformed into the average-case Gaussian-robustness by "smoothing" a classifier, i.e., by considering the averaged prediction over Gaussian noise. In this paradigm, one should rethink the notion of adversarial robustness in terms of generalization ability of a classifier under noisy observations. We found that the trade-off between accuracy and certified robustness of smoothed classifiers can be greatly controlled by simply regularizing the prediction consistency over noise. This relationship allows us to design a robust training objective without approximating a non-existing smoothed classifier, e.g., via soft smoothing. Our experiments under various deep neural network architectures and datasets show that the "certified" ℓ 2 -robustness can be dramatically improved with the proposed regularization, even achieving better or comparable results to the state-of-the-art approaches with significantly less training costs and hyperparameters. | - |
dc.language | English | - |
dc.publisher | NeurIPS committee | - |
dc.title | Consistency Regularization for Certified Robustness of Smoothed Classifiers | - |
dc.type | Conference | - |
dc.identifier.scopusid | 2-s2.0-85108401729 | - |
dc.type.rims | CONF | - |
dc.citation.publicationname | 34th Conference on Neural Information Processing Systems (NeurIPS) 2020 | - |
dc.identifier.conferencecountry | CN | - |
dc.identifier.conferencelocation | Virtual | - |
dc.contributor.localauthor | Shin, Jinwoo | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.