DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Kim, Junmo | - |
dc.contributor.advisor | 김준모 | - |
dc.contributor.author | Yim, Junho | - |
dc.date.accessioned | 2019-08-25T02:44:35Z | - |
dc.date.available | 2019-08-25T02:44:35Z | - |
dc.date.issued | 2019 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=842386&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/265164 | - |
dc.description | 학위논문(박사) - 한국과학기술원 : 전기및전자공학부, 2019.2,[vii, 63 p. :] | - |
dc.description.abstract | In the area of deep neural networks (DNNs), there are several research directions to improve the network performance, such as research about the network architecture, optimization techniques, etc. For a part of these various research ways, we focus on multitask learning techniques. Although DNNs are generally trained to minimize the loss between output and label, we propose another loss term that helps generate high performance in the main task. By using the auxiliary loss term that makes DNNs to generate input images, proposed model achieves high performance better than DNNs with general single loss term in face recognition task. Furthermore, we proposed a novel knowledge distillation technique which makes another loss term that minimizes the distances between the features of the main DNN and the other DNN. In several image classification datasets, we showed that our proposed multitask learning model generates high performance better than the general single loss model. We conducted further research on not only the supervised learning but also unsupervised learning. In the generative adversarial network (GAN) framework which is widely used to generate specific distributions, we proposed novel task that creates an intermediate domain space from two existing domains by adding one more auxiliary loss term at the general GAN loss terms. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | Multitask learning▼adeep neural network▼agenerative adversarial network▼aface recognition▼aimage classification | - |
dc.subject | 멀티테스크 학습▼a깊은 신경망▼a생성적 적대 신경망▼a얼굴 인식▼a지식 추출 기법▼a물체 인식 | - |
dc.title | Deep learning methods and architectures using multi-task learning | - |
dc.title.alternative | 멀티테스크 러닝을 이용한 딥러닝 기법 및 구조 : 성능향상 기법 및 이미지 창조 | - |
dc.type | Thesis(Ph.D) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :전기및전자공학부, | - |
dc.contributor.alternativeauthor | 임준호 | - |
dc.title.subtitle | Performance Improvement and Image Creation | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.