Jun Zhu

Project Scientist

8116 Gates-Hillman Center (GHC)

Machine Learning Department

Carnegie Mellon University

Pittsburgh, PA 15213

E-mail: junzhu@cs [dot] cmu [dot] edu

I'm a post-doctoral fellow at the Machine Learning Department in CMU, working with Prof. Eric Xing in the Sailing Lab. My current research interests are focusing on statistical machine learning with applications to computer vision, text mining, and web mining. Please see my recent projects and publications for more details.

• Research Interests (CV)
  • Statistical Machine Learning, Probabilistic Graphical Models, Large-margin Learning, Bayesian Nonparametrics
  • Applications in Text Mining, Web Mining, Computer Vision, Social Media Analysis, etc.
• Visiting Experience
  • Visiting Researcher at Stanford Vision Group, 2010.2~3 & 2010.10~11. Worked with Prof. Li Fei-Fei
  • Visiting Researcher at Machine Learning Department, CMU, 2007 & 2008. Worked with Prof. Eric Xing
  • Research Assistant at Microsoft Research Asia, 2004 ~ 2007 & 2008. Worked with Dr. Zaiqing Nie, Dr. Ji-Rong Wen, Dr. Lei Zhang, and Dr. Wei-Ying Ma
• Projects & Publications (full list ...)
  • Joint max-entropy and large-margin learning: We present Maximum Entropy Discrimination Markov Networks (MaxEntNet) for learning structured output prediction models. In its full generality MaxEntNet offers a number of advantages (more ...)
  • Learning predictive latent subspace representations: We propose a large-margin learning framework for discovering latent representations that are predictive and interpretable (See MedLDA & Infinite SVM). We also do Structured Prediction Learning and Relational Learning in the presence of latent variables (more ...)
  • Sparse Learning in High Dimensions: We develop efficient algorithms for Structure Learning of Markov networks and feature selection. We also develop efficient Sparse Coding methods for discovering topical representations with rich features (more ...)
  • Applications: I'm working on Web Mining, Social Media Analysis, Text Mining, and Computer Vision problems. Our technique of StatSnowball on extracting web entity relationships have been successfully used in EntityCube and its Chinese version Renlifang entity relationship search engines (more ...)
• Scientific Software
  • C++ implementation of MedLDA: max-margin supervised topic models
  • C++ implementation of Infinite SVM and Infinite Latent SVM coming soon
  • C++ implementation of Sparse Topical Coding and Conditional Topical Coding
• Academic Activities
  • Conference PC Member: NIPS'11, ICML'11, SIGKDD'11, EMNLP'11, ICDE'11, ICML'10, NIPS'10, SIGKDD'10, AAAI'10, ACL'10, EMNLP'10.
  • Journal Reviewer: IEEE Trans. on PAMI, Machine Learning Journal, Journal of Artificial Intelligence Research, ACM Trans. on Knowledge Discovery from Data.
  • Student Volunteer: ICML'05, 08 & 09; SIGKDD'06, 07 & 09; NIPS'08.
• Teaching Assistant
  • CMU Course: 10-708, Probabilistic Graphical Models, Fall 2009.
  • Dragon Star Course: Advanced Machine Learning, Summer 2009, Beijing.

• Blog (Chinese)
  
• Friends and Collaborators
  • Fan Guo, Ni Lao, Chong Wang, Le Zhao, Lei Li, Xiaolin Shi, Lei Wu, Amr Ahmed, Zaiqing Nie, Ji-Rong Wen, Lei Zhang