Publications
Detailed Information
Bayesian Prediction and Regression from Visual Data : 영상 데이터에 대한 베이지안 예측 및 회귀 분석
DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | 최진영 | - |
dc.contributor.author | 유영준 | - |
dc.date.accessioned | 2017-07-13T07:20:48Z | - |
dc.date.available | 2017-07-13T07:20:48Z | - |
dc.date.issued | 2017-02 | - |
dc.identifier.other | 000000141832 | - |
dc.identifier.uri | https://hdl.handle.net/10371/119274 | - |
dc.description | 학위논문 (박사)-- 서울대학교 대학원 : 전기·컴퓨터공학부, 2017. 2. 최진영. | - |
dc.description.abstract | This dissertation proposes a new high dimensional regression / prediction method for
diverse visual data pairs. In contrast to other regression / prediction methods, the proposed method focuses on the case where output responses are on a complex high dimensional manifold, such as images. In handling the complex data, the latent space embedding the information of the data is used for efficient regression / prediction. The dimensionality reduction methods into the latent space and the regression/prediction methods are designed as a Bayesian framework. For the prediction problem, the dissertation proposes a method to extract latent semantics on motion dynamics given in visual sequences. To this end, a Bayesian inference model is developed to capture the regional and temporal semantics of the dynamics data. The proposed Bayesian model is a hierarchical fusion of Gaussian mixture model and topic mixture model. It finds regional pattern information through topic mixture model and derives temporal co-occurrence of regional patterns through Gaussian mixture model. To infer the proposed model, the dissertation proposes a new sampling method that enables efficient inference. For the regression problem, we propose a method that makes a regression in the latent space for general and complex visual data pairs. This allows the latent space to imply the essential properties of the data pairs required for regression. For the purpose, a regression model is designed so that the regression in latent space should coincide with the regression in data space. The whole models are designed as Bayesian framework, and inferred by variational autoencoder framework. | - |
dc.description.tableofcontents | Chapter 1 Introduction 1
1.1 Background 1 1.2 Related Work 4 1.3 Contents of Research 7 1.4 Thesis Organization 8 Chapter 2 Preliminaries 11 2.1 Overview 11 2.2 Bayesian Statistics for Generative Model 11 2.2.1 Overview 11 2.2.2 Bayes Theorem 12 2.2.3 Example: Bayesian Curve Fitting Problem 13 2.2.4 Bayesian Model Comparison 15 2.2.5 Approximate Inference 18 2.3 Sampling Based Methods 19 2.3.1 Overview 19 2.3.2 Monte Carlo Method 19 2.3.3 Basic Sampling Methods 20 2.3.4 Markov Chain Monte Carlo 24 2.3.5 Gibbs Sampling 26 2.4 Optimization Based Methods 27 2.4.1 Overview 27 2.4.2 Kullback-Leibler Divergence 28 2.4.3 Variational Inference 28 2.4.4 Mean-Field Approximation 29 2.4.5 Autoencoding Variational Bayes 31 2.5 Gaussian Process Regression 33 2.5.1 Overview 33 2.5.2 Weighted Space View 33 2.5.3 Function Space View 34 Chapter 3 Prediction from Visual Data 37 3.1 Overall Scheme 37 3.2 Conversion of Input Trajectories 40 3.3 Hierarchical Topic-Gaussian Mixture Model 40 3.4 Inference of the HTGMM 44 3.5 Deterministic Method for Path Prediction 50 Chapter 4 Regression of Visual Data 55 4.1 Overall Scheme 55 4.2 Variational Autoencoded Regression 59 4.3 Model Description 61 4.4 Training 63 4.5 Implementation Detail 67 Chapter 5 Experiments 69 5.1 Visual Prediction 69 5.1.1 Dataset 69 5.1.2 Comparison Methods 70 5.1.3 Qualitative Evaluation 71 5.1.4 Quantitative Evaluation 80 5.1.5 Summary 81 5.2 Visual Regression 82 5.2.1 Dataset 82 5.2.2 Sports Data Sequences 82 5.2.3 Human Pose Reconstruction 95 5.2.4 Summary 100 Chapter 6 Conclusion 103 6.1 Contribution 103 6.2 Future work 104 Bibliography 105 초록 119 | - |
dc.format | application/pdf | - |
dc.format.extent | 12181174 bytes | - |
dc.format.medium | application/pdf | - |
dc.language.iso | en | - |
dc.publisher | 서울대학교 대학원 | - |
dc.subject | 회귀분석 | - |
dc.subject | 확률적 그래프 모델 | - |
dc.subject | 계층적 생성 모델 | - |
dc.subject | 근사 추론 | - |
dc.subject.ddc | 621 | - |
dc.title | Bayesian Prediction and Regression from Visual Data | - |
dc.title.alternative | 영상 데이터에 대한 베이지안 예측 및 회귀 분석 | - |
dc.type | Thesis | - |
dc.contributor.AlternativeAuthor | YoungJoon Yoo | - |
dc.description.degree | Doctor | - |
dc.citation.pages | 119 | - |
dc.contributor.affiliation | 공과대학 전기·컴퓨터공학부 | - |
dc.date.awarded | 2017-02 | - |
- Appears in Collections:
- Files in This Item:
Item View & Download Count
Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.