DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Lee, Juho | - |
dc.contributor.advisor | 이주호 | - |
dc.contributor.author | Jang, Sunguk | - |
dc.date.accessioned | 2023-06-22T19:31:20Z | - |
dc.date.available | 2023-06-22T19:31:20Z | - |
dc.date.issued | 2023 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=1032329&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/308203 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 김재철AI대학원, 2023.2,[iv, 34 p. :] | - |
dc.description.abstract | Decoupling representation learning and classifier learning has been shown to be effective in classification with long-tailed data. There are two main ingredients in constructing a decoupled learning scheme | - |
dc.description.abstract | 1) how to train the feature extractor for representation learning so that it provides generalizable representations and 2) how to re-train the classifier that constructs proper decision boundaries by handling class imbalances in long-tailed data. In this work, we first apply Stochastic Weight Averaging (SWA), an optimization technique for improving the generalization of deep neural networks, to obtain better generalizing feature extractors for long-tailed classification. We then propose a novel classifier re-training algorithm based on stochastic representation obtained from the SWA-Gaussian, a Gaussian perturbed SWA, and a self-distillation strategy that can harness the diverse stochastic representations based on uncertainty estimates to build more robust classifiers. Extensive experiments on several dataset benchmarks show that our proposed method improves upon previous methods both in terms of prediction accuracy and uncertainty estimation. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | Long-tailed classification▼aDecoupled learning▼aStochastic weight averaging▼aStochastic representations | - |
dc.subject | 긴 꼬리 분류▼a분리 학습▼a확률적 가중치 평균화▼a확률적 표현 | - |
dc.title | Decoupled training for long-tailed classification with stochastic representations | - |
dc.title.alternative | 확률적 표현을 활용한 긴 꼬리 분류를 위한 분리 학습 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :김재철AI대학원, | - |
dc.contributor.alternativeauthor | 장성욱 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.