서브메뉴
검색
Learning Representations Toward the Understanding of Out-Of-Distribution for Neural Networks
Learning Representations Toward the Understanding of Out-Of-Distribution for Neural Networks
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105541
- ISBN
- 9798263395490
- DDC
- 388.312
- 저자명
- Kwon, Gukyeong.
- 서명/저자
- Learning Representations Toward the Understanding of Out-Of-Distribution for Neural Networks
- 발행사항
- [Sl] : Georgia Institute of Technology, 2021
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2021
- 형태사항
- 125 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-05, Section: A.
- 주기사항
- Advisor: Alregib, Ghassan.
- 학위논문주기
- Thesis (Ph.D.)--Georgia Institute of Technology, 2021.
- 초록/해제
- 요약Data-driven representations achieve powerful generalization performance in diverse information processing tasks. However, the generalization is often limited to test data from the same distribution as training data (in-distribution (ID)). In addition, the neural networks often make overconfident and incorrect predictions for data outside training distribution, called out-of-distribution (OOD). In this dissertation, we develop representations that can characterize OOD for the neural networks and utilize the characterization to efficiently generalize to OOD.We categorize the data-driven representations based on information flow in neural networks and develop novel gradient-based representations. In particular, we utilize the backpropagated gradients to represent what the neural networks has not learned in the data. The capability of gradient-based representations for OOD characterization is comprehensively analyzed in comparison with standard activation-based representations. We also develop a regularization technique for the gradient-based representations to better characterize OOD. We develop an anomaly detection algorithm named GradCon using the gradient constraint and achieve state-of-the-art performance in OOD detection.We also propose activation-based representations learned with auxiliary information to efficiently generalize to data from OOD. We use an unsupervised learning framework to learn the aligned representations of visual and attribute data. This aligned representation are utilized to calibrate the overconfident prediction toward ID classes. The generalization performance of the aligned representations is validated in the application of generalized zero-shot learning. Our developed GZSL method, GatingAE, achieves state-of-the-art performance in generalizing to OOD without using labeled OOD data. Also, balanced performance for both ID and OOD is achieved by mitigating the prediction bias presented in the network. Finally, GatingAE requires significantly less number of model parameters compared to other state-of-the-art methods.
- 일반주제명
- Automobile parking
- 일반주제명
- Wavelet transforms
- 일반주제명
- Back propagation
- 일반주제명
- Fourier transforms
- 일반주제명
- Street signs
- 일반주제명
- Neural networks
- 일반주제명
- Traffic control
- 일반주제명
- Information processing
- 일반주제명
- Data compression
- 일반주제명
- Computer science
- 일반주제명
- Mathematics
- 일반주제명
- Transportation
- 일반주제명
- Urban planning
- 기본자료저록
- Dissertations Abstracts International. 87-05A.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2021 us c eng d■001000017360524
■00520260202105541
■006m o d
■007cr#unu||||||||
■020 ▼a9798263395490
■035 ▼a(MiAaPQ)AAI32315129
■035 ▼a(MiAaPQ)GeorgiaTech65045
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a388.312
■1001 ▼aKwon, Gukyeong.
■24510▼aLearning Representations Toward the Understanding of Out-Of-Distribution for Neural Networks
■260 ▼a[Sl]▼bGeorgia Institute of Technology▼c2021
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2021
■300 ▼a125 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-05, Section: A.
■500 ▼aAdvisor: Alregib, Ghassan.
■5021 ▼aThesis (Ph.D.)--Georgia Institute of Technology, 2021.
■520 ▼aData-driven representations achieve powerful generalization performance in diverse information processing tasks. However, the generalization is often limited to test data from the same distribution as training data (in-distribution (ID)). In addition, the neural networks often make overconfident and incorrect predictions for data outside training distribution, called out-of-distribution (OOD). In this dissertation, we develop representations that can characterize OOD for the neural networks and utilize the characterization to efficiently generalize to OOD.We categorize the data-driven representations based on information flow in neural networks and develop novel gradient-based representations. In particular, we utilize the backpropagated gradients to represent what the neural networks has not learned in the data. The capability of gradient-based representations for OOD characterization is comprehensively analyzed in comparison with standard activation-based representations. We also develop a regularization technique for the gradient-based representations to better characterize OOD. We develop an anomaly detection algorithm named GradCon using the gradient constraint and achieve state-of-the-art performance in OOD detection.We also propose activation-based representations learned with auxiliary information to efficiently generalize to data from OOD. We use an unsupervised learning framework to learn the aligned representations of visual and attribute data. This aligned representation are utilized to calibrate the overconfident prediction toward ID classes. The generalization performance of the aligned representations is validated in the application of generalized zero-shot learning. Our developed GZSL method, GatingAE, achieves state-of-the-art performance in generalizing to OOD without using labeled OOD data. Also, balanced performance for both ID and OOD is achieved by mitigating the prediction bias presented in the network. Finally, GatingAE requires significantly less number of model parameters compared to other state-of-the-art methods.
■590 ▼aSchool code: 0078.
■650 4▼aAutomobile parking
■650 4▼aWavelet transforms
■650 4▼aBack propagation
■650 4▼aFourier transforms
■650 4▼aStreet signs
■650 4▼aNeural networks
■650 4▼aTraffic control
■650 4▼aInformation processing
■650 4▼aData compression
■650 4▼aComputer science
■650 4▼aMathematics
■650 4▼aTransportation
■650 4▼aUrban planning
■690 ▼a0800
■690 ▼a0984
■690 ▼a0405
■690 ▼a0709
■690 ▼a0999
■71020▼aGeorgia Institute of Technology.
■7730 ▼tDissertations Abstracts International▼g87-05A.
■790 ▼a0078
■791 ▼aPh.D.
■792 ▼a2021
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17360524▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


