Advanced Probabilistic Modeling Algorithms for Clustering ...
- 1. CV – Thesis Outline Shipeng Yu Shipeng Yu Ph.D. candidate Institute for Computer Science, University of Munich, Germany Siemens Corporate Technology, Munich, Germany CONTACT INFORMATION Tel: +49-89-636-47926 (O), +49-89-8898-1948 (H), +49-176-2400-1886 (Mobile) Fax: +49-89-636-45456 (O) URL: http://www.dbs.ifi.lmu.de/~spyu/ Email: spyu@dbs.ifi.lmu.de, shipeng.yu.ext@siemens.com, yushipeng@yahoo.com Address: Gerhart-Hauptmann-Ring 3, D-81737, Munich, Germany RESEARCH INTERESTS Statistical Machine Learning, Probabilistic Modeling, Bayesian Data Analysis Bayesian Clustering, Dimensionality Reduction, Supervised Projection Document Modeling, Collaborative Filtering, Ranking Information Retrieval and Extraction, Text Mining, Web Search and Mining EDUCATION Ph.D. in Institute for Computer Science, University of Munich (9/2003 - exp. 9/2006) Supervisor: Prof. Dr. Hans-Peter Kriegel, Dr. Volker Tresp M.Sc. in Information Science, School of Mathematical Sciences, Peking University, China (9/2000 - 7/2003) Supervisor: Prof. Dr. Zuoquan Lin GPA: 91.6/100 Rank: 2/70 B.Sc. in Information Science, School of Mathematical Sciences, Peking University, China (9/1996 - 1/10
- 2. CV – Thesis Outline Shipeng Yu 7/2000) GPA: 93.5/100 for major, 91.7/100 for all Rank: 2/41 RESEARCH EXPERIENCE 9/2003 - Present: Guest Research Scientist at Siemens Corporate Technology, Munich, Germany Directions: statistical machine learning, data mining and information retrieval Proposed a Dirichlet enhanced latent semantic analysis for documents Introduced a variational Bayesian method for infinite mixture of exponential family distributions Co-invented a supervised projection model and its probabilistic extensions and interpretations Co-invented a soft clustering algorithm on graph with applications on semi-supervised learning Introduced a multi-task Gaussian Process model for collaborative ranking problem Proposed a supervised and a semi-supervised PCA which can be applied to very large data sets 7/2005 - 9/2005: Research Internship at Microsoft Research Cambridge, Cambridge, UK Topic: Bayesian web ranking with TrueSkillTM System Addressed web ranking problem by extending TrueSkillTM system for games Considered time-independent ranking and feature-dependent ranking for web ranking scenario Implemented the whole algorithm to handle millions of document-query pairs Proposed a probabilistic method to evaluate judges and measure query difficulties 9/2001 - 8/2003: Visiting Student at Microsoft Research Asia, Beijing, China Directions: web page segmentation and applications on web information retrieval, extraction and mining Proposed a novel vision-based content structuring method for web page Verified that vision-based page segmentation (VIPS) could significantly improve web IR performance Proposed a combined page segmentation approach which achieved the best performance Introduced a new strategy for automatic web information extraction based on web content structure 1/2000 - 7/2003: Research Assistant at PKU-Goldtech Networking R&D Center, Beijing, China Directions: web-based DRP systems with security issues 2/10
- 3. CV – Thesis Outline Shipeng Yu Team Member of WebDaemon, a role-based access control (RBAC) system on the web Project Leader of TYGXC, an Internet-based DRP system for TianYin Corporation, China Team Member of TCLJXC, an Internet-based DRP system for TCL Corporation, China SELECTED PUBLICATIONS Shipeng Yu, Kai Yu, Mingrui Wu, Volker Tresp, Hans-Peter Kriegel. Supervised Probabilistic Principal Component Analysis. To appear in the 12th ACM International Conference on Knowledge Discovery and Data Mining (KDD’2006), Philadelphia, PA, 2006. Shipeng Yu, Kai Yu, Volker Tresp, Hans-Peter Kriegel. Collaborative Ordinal Regression. To appear in the 23rd International Conference on Machine Learning (ICML'2006), Pittsburgh, PA, 2006. Shipeng Yu, Kai Yu, Volker Tresp. Collaborative Ordinal Regression. In the Proceedings of Workshop on Learning to Rank at Neural Information Processing Systems 18 (NIPS'2005), Whistler, Canada, 2005. Kai Yu, Shipeng Yu, Volker Tresp. Soft Clustering on Graphs. Advances in Neural Information Processing Systems 18 (NIPS'2005), Vancouver, Canada, 2005. Yi Huang, Kai Yu, Matthias Schubert, Shipeng Yu, Hans-Peter Kriegel. Hierarchy-Regularized Latent Semantic Indexing. In the 5th IEEE International Conference on Data Mining (ICDM'2005), New Orleans, Louisiana, 2005. Shipeng Yu, Kai Yu, Volker Tresp, Hans-Peter Kriegel. A Probabilistic Clustering-Projection Model for Discrete Data. In the 9th European Conference on Principles and Practice of Knowledge Discovery in Databases (PKDD'2005), Porto, Portugal, 2005. (Best Paper Runner-Up / Best Student Paper Runner-Up) Kai Yu, Shipeng Yu, Volker Tresp. Blockwise Supervised Inference on Large Graphs. In the Proceedings of Workshop on Learning with Partially Classified Training Data at the 22nd International Conference on Machine Learning (ICML'2005), Bonn, Germany, 2005. Zhao Xu, Volker Tresp, Kai Yu, Shipeng Yu, Hans-Peter Kriegel. Dirichlet Enhanced Relational Learning. In the 22nd International Conference on Machine Learning (ICML'2005), Bonn, Germany, 2005. Kai Yu, Shipeng Yu, Volker Tresp. Multi-Label Informed Latent Semantic Indexing. In the 28th Annual International ACM SIGIR Conference (SIGIR'2005), Salvador, Brazil, 2005. 3/10
- 4. CV – Thesis Outline Shipeng Yu Kai Yu, Shipeng Yu, Volker Tresp. Multi-Output Regularized Projection. In the International IEEE Conference on Computer Vision and Pattern Recognition (CVPR’2005), San Diego, CA, 2005. Kai Yu, Shipeng Yu, Volker Tresp. Dirichlet Enhanced Latent Semantic Analysis. In the Tenth International Workshop on Artificial Intelligence and Statistics (AISTATS'2005), Barbados, 2005. Kai Yu, Volker Tresp, Shipeng Yu. A Nonparametric Hierarchical Bayesian Framework for Information Filtering. In the 27th Annual International ACM SIGIR Conference (SIGIR'2004), Sheffield, UK, 2004. Deng Cai, Shipeng Yu, Ji-Rong Wen, Wei-Ying Ma. Block-based Web Search. In the 27th Annual International ACM SIGIR Conference (SIGIR'2004), Sheffield, UK, 2004. Ji-Rong Wen, Ruihua Song, Deng Cai, Kaihua Zhu, Shipeng Yu, Shaozhi Ye, Wei-Ying Ma. Microsoft Research Asia at the Web Track of TREC 2003, In the Twelfth Text Retrieval Conference (TREC'12), 2003. Deng Cai, Shipeng Yu, Ji-Rong Wen, Wei-Ying Ma. VIPS: A Vision-based Page Segmentation Algorithm. Microsoft Technical Report, MSR-TR-2003-79, 2003. Deng Cai, Shipeng Yu, Ji-Rong Wen, Wei-Ying Ma. Extracting Content Structure for Web Pages based on Visual Representation. In the Fifth Asia Pacific Web Conference (APWeb’2003), Xi’an, China, 2003. Shipeng Yu, Deng Cai, Ji-Rong Wen, Wei-Ying Ma. Improving Pseudo-Relevance Feedback in Web Information Retrieval Using Web Page Segmentation. In the Twelfth International World Wide Web Conference (WWW’2003), Budapest, Hungary, 2003. TALKS Bayesian Web Ranking for Information Retrieval. Microsoft Research Cambridge, UK, Mar. 2006. Collaborative Ordinal Regression. NIPS’05 Workshop on Learning to Rank, Dec. 2005. A Probabilistic Clustering-Projection Model for Discrete Data. The 9th European Conference on Principles and Practice of Knowledge Discovery in Databases, Porto, Portugal, Oct. 2005. TrueSkillTM Hits the Web: Bayesian Ranking Methods for Web Search. Intern Talk, Microsoft Research Cambridge, UK, Sep. 2005. Multi-Label Informed Latent Semantic Indexing. The 28th Annual International ACM SIGIR 4/10
- 5. CV – Thesis Outline Shipeng Yu Conference, Salvador, Brazil, Aug. 2005. Dirichlet Enhanced Latent Semantic Analysis. Lernen, Wissensentdeckung und Adaptivität (LWA’04), Humboldt University, Berlin, Germany, Oct. 2004. Enhancing Web Information Retrieval Using Web Page Segmentation. Siemens Corporate Technology, Munich, Germany, Nov. 2003. Enhancing Web Information Retrieval Using Web Page Segmentation. Institute for Informatics, University of Munich, Munich, Germany, Oct. 2003. TEACHING EXPERIENCE 9/2000 - 1/2002: Lecturer in School of Continuing Education, Peking University Assembly Language, Autumn 2000 and Autumn 2001 9/2001 - 1/2003: Teaching Assistant in School of Mathematical Sciences, Peking University Introduction to Database Systems, Autumn 2002 MATLAB System, Spring 2002 Statistical Computation, Autumn 2001 PATENTS Collaborative Ordinal Regression. Filed November 2005, Siemens Corporate Technology. Soft Clustering on Graphs. Filed October 2005, Siemens Corporate Technology. Multi-Output Regularized Projection. Filed June 2005, Siemens Corporate Technology. Dirichlet Enhanced Latent Semantic Analysis. Filed October 2004, Siemens Corporate Technology. Vision-Based Document Segmentation. Filed July 2003, Microsoft Research Asia. AWARDS AND HONORS Student Travel Grant for ICML’2006 Best Paper Runner-Up & Best Student Paper Runner-Up Award in PKDD’2005 Student Travel Grant for SIGIR’2005 5/10
- 6. CV – Thesis Outline Shipeng Yu Student Travel Grant for ICML’2005 Attendee of Machine Learning Summer School (MLSS’2004), Berder, France, Sep. 2004 Student Travel Grant for SIGIR’2004 Siemens Doctoral Scholarship, Sep. 2003 - Aug. 2006 Outstanding Undergraduate Student of Beijing (<1%), 2000 Outstanding Undergraduate Student of Peking University (<2%), 2000 First Prize in National Mathematical Modeling Competition, 1999 Award for Excellent Student of Peking University, in three successive years 1997 - 1999 Scholarships, in three successive years 1997 - 1999 Rank 3 (4 is the best) in National Computer Rank Examination Certificate, 1998 First Prize in National Mathematics Competition for High School Students, 1995 COMPUTER SKILLS Languages mastered: C, C++, Java, MATLAB, .NET, Perl, Maple Systems experienced: MS Windows, Linux OTHERS Fluent in Chinese and English. Read German and French Master accordion. Degree 5 (out of 10) in the National Accordion Certificate Enjoy badminton, table tennis, swimming, traveling and reading REFERENCES Available upon request. 6/10
- 7. CV – Thesis Outline Shipeng Yu Advanced Probabilistic Modeling Algorithms for Clustering, Projection and Ranking Abstract List of Algorithms List of Figures List of Tables Outline 1. Introduction a) Unsupervised Learning i. Clustering Models ii. Projection Models b) Supervised Learning i. Ranking Models c) Probabilistic Modeling Framework i. Graphical Models ii. Inference and Learning 2. Probabilistic Models for Clustering a) Introduction b) Finite Mixture Modeling for Clustering i. Exponential Family Distributions and the Conjugate Family ii. Mixture of Exponential Family Distributions iii. Model Inference and Parameter Estimation iv. Empirical Studies 7/10
- 8. CV – Thesis Outline Shipeng Yu c) Infinite Mixture Models i. Infinite Mixture Models ii. Dirichlet Processes iii. Inference and Parameter Estimation iv. Empirical Studies 3. Probabilistic Models for Projection a) Introduction b) Probabilistic Projection for Non-linear PCA i. Principal Component Analysis (PCA) ii. Kernel Principal Component Analysis (Kernel PCA) iii. Probabilistic PCA (PPCA) iv. Probabilistic Kernel PCA Model v. Incremental Kernel PCA vi. Empirical Studies c) Supervised Feature Projection i. Supervised Latent Variable Model ii. Multi-Output Regularized Feature Projection iii. Connections to Related Works iv. Discussions v. Empirical Studies d) Supervised Probabilistic Principal Component Analysis 8/10
- 9. CV – Thesis Outline Shipeng Yu i. Supervised Probabilistic PCA (SPPCA) ii. Semi-Supervised Probabilistic PCA (S2PPCA) iii. Learning in SPPCA Models iv. Theoretical Justification v. Empirical Studies 4. Probabilistic Models for Joint Projection-Clustering Models a) Introduction b) Probabilistic Projection-Clustering Models for Discrete Data i. Probabilistic Framework ii. Model Learning and Inference iii. Empirical Studies c) Probabilistic Projection-Clustering Models for Continuous Data i. Probabilistic Framework ii. Model Learning and Inference iii. Empirical Studies 5. Probabilistic Models for Ranking a) Introduction b) Bayesian Ranking Models c) Collaborative Ranking Models i. Gaussian Processes for Multi-task Learning ii. Collaborative Ordinal Regression 9/10
- 10. CV – Thesis Outline Shipeng Yu iii. Empirical Studies 6. Conclusion and Future Work a) Summary b) Future works Appendix A. Learning in Probabilistic Models a) MCMC b) Expectation-Maximization (EM) c) Variational Bayes d) Expectation-Propagation Index Reference 10/10