DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Kim, Lee-Sup | - |
dc.contributor.advisor | 김이섭 | - |
dc.contributor.author | Ra, Jongwoo | - |
dc.date.accessioned | 2023-06-26T19:33:40Z | - |
dc.date.available | 2023-06-26T19:33:40Z | - |
dc.date.issued | 2023 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=1033102&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/309838 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 전기및전자공학부, 2023.2,[iv, 35 p. :] | - |
dc.description.abstract | The robustness of neural networks has been emphasized since some works discovered that small perturbations could cause misclassification. Various defense methods against adversarial attacks have been proposed, and adversarial training is one of the most effective. Adversarial training generates adversarial examples and learns such data so that models can classify adversarial examples accurately. Multiple steps of forward pass and backpropagation are necessary to generate adversarial examples, and it consumes enormous computation costs and latency. This work accelerates adversarial training by selective computation of features of which gradients are large. This work also proposes a fast gradient top-k method exploiting the distribution of exponents and selectively computes large gradients in FP32. Efficient hardware to support selective computation, exponent approximation, and gradient top-k is designed with a small overhead. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | CNN▼aAdversarial attack▼aAdversarial Training▼aSelective activation computation▼aSelective gradient computation | - |
dc.subject | 컨볼루셔널 뉴럴 네트워크▼a적대적 공격▼a적대적 훈련▼a선택적 활성화 연산▼a선택적 그래디언트 연산 | - |
dc.title | Algorithm / hardware co-design for adversarial training acceleration through selective computations | - |
dc.title.alternative | 적대적 학습의 가속을 위한 선택적 연산 알고리즘 및 하드웨어 공동설계 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :전기및전자공학부, | - |
dc.contributor.alternativeauthor | 라종우 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.