DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Hwang, Sung Ju | - |
dc.contributor.advisor | 황성주 | - |
dc.contributor.author | Madaan, Divyam | - |
dc.date.accessioned | 2022-04-27T19:32:12Z | - |
dc.date.available | 2022-04-27T19:32:12Z | - |
dc.date.issued | 2021 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=963377&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/296160 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 전산학부, 2021.8,[iv, 35 p. :] | - |
dc.description.abstract | Despite the remarkable performance of deep neural networks on various computer vision tasks, they are known to be susceptible to adversarial perturbations, which makes it challenging to deploy them in real-world safety-critical applications. In this thesis, we conjecture that the leading cause of the adversarial vulnerability is the distortion in the latent feature space and provide methods to suppress them effectively. We propose a Bayesian framework to prune features with high vulnerability to reduce vulnerability and loss on adversarial samples. We validate our Adversarial Neural Pruning with Vulnerability Suppression (ANP-VS) method on multiple benchmark datasets. It obtains state-of-the-art adversarial robustness and improves the performance on clean examples, using only a fraction of the parameters used by the complete network. We further propose a novel meta-learning framework that explicitly learns to generate noise to improve the model’s robustness against multiple types of attacks. Its key component is Meta Noise Generator (MNG) that outputs optimal noise to stochastically perturb a given sample, such that it helps lower the error on diverse adversarial perturbations. We validate the robustness of models trained by our scheme on various datasets and against a wide variety of perturbations, demonstrating that it significantly outperforms the baselines across multiple perturbations with a marginal computational cost. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | Adversarial defense▼aNetwork pruning▼aMeta-learning | - |
dc.subject | 적대적 방어▼a신경망 프루닝▼a메타 러닝 | - |
dc.title | Generalizable robust deep learning via adversarial pruning and meta-noise generation | - |
dc.title.alternative | 적대적 프루닝과 메타 노이즈 생성 기반 일반화 가능한 강건한 딥러닝 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :전산학부, | - |
dc.contributor.alternativeauthor | 마단디뱜 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.