" Only in silence the word, only in dark the light, only in dying life: bright the hawk's flight on the empty sky "
&mdash Ursula K. Le Guin (A Wizard of Earthsea)

Qinfeng (Javen) Shi

PhD Candidate (since July 2006)

Institutions: Contact:
  • T: +61-2-6267-6331
  • M: +61-4-2370-3886
  • Email: qinfeng.shi at ieee.org

CV:

[pdf]

Research Interests

Machine Learning, Compressive Sensing, Image and Video Analysis;
Particularly Structured Model Learning, Kernel Methods, PAC-Bayes Bound and Compressibility Analysis.
Current Research: Structured estimation often involves exponential many possible labels or configurations. How to efficiently infer and learn such models remain a challenging and fun task. Assessing the performance in theory is a joy.
Past Research: Synthetic Aperture Radar (SAR) images taken from airborne or space platforms can have high spatial resolution, but are affected by the speckle phenomenon which makes the extraction of useful information a difficult task. My master thesis focused on SAR image denoising, edge detection,and segmentation.

My Supervisors:

Alex J. Smola (Now in Yahoo! Research), S. V. N. Vishwanathan (Now in Purdue), Li Cheng (Now in TTI)
Tiberio Caetano, Richard Hartley

Publications

11 Qinfeng Shi, Mark Reid, Tiberio Caetano, Hybrid model of Conditional Random Field and Support Vector Machine, Workshop at the 23rd Annual Conference on Neural Information Processing Systems, Canada, Dec. 2009. [ pdf]
10 Qinfeng Shi, James Petterson, Gideon Dror, John Langford, Alex Smola and Vishy Vishwanathan, Hash Kernels for Structured Data, Journal of Machine Learning Research, Nov. 2009. [ pdf]
9 Qinfeng Shi, Li Cheng, Luping Zhou and Dale Schuurmans, Discriminative Maximum Margin Image Object Categorization with Exact Inference, The 5th International Conference on Image and Graphics, Xi'an, Sep 20-23, 2009. [ pdf]
8 Qinfeng Shi, James Petterson, Gideon Dror, John Langford, Alex Smola, Alex Strehl and Vishy Vishwanathan, Hash Kernels, Twelfth International Conference on Artificial Intelligence and Statistics, Florida, Apirl 14-19, 2009. [pdf]
7 Qinfeng Shi, Li Wang, Li Cheng and Alex Smola, Discriminative Human Action Segmentation and Recognition using Semi-Markov Model (long version with BMRM and PAC-Bayes Bound), International Journal of Computer Vision, Accepted under Revision. 2009. [pdf]
6 Qinfeng Shi, Li Wang, Li Cheng and Alex Smola, Discriminative Human Action Segmentation and Recognition using Semi-Markov Model (short version with cutting plane method), In IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR 08),Anchorage, Alaska, June 23-28, 2008. [pdf]
5 Qinfeng Shi, Yasemin Altun , Alex Smola and S.V.N. Vishwanathan, Automatic Paragraph Segmentation via Max-Margin Semi-Markov Models, In Proceedings of the 2007 Conference on Empirical Methods in Natural Language Processing (EMNLP-CoNLL07) , Jun 2007, pp. 640-648. [pdf]
4 Ying Li, Qinfeng Shi, Yanning Zhang and Rongchun Zhao, A study on Automated segmentation algorithm of SAR Image, Journal of Electronics & Information Technology (Chinese), 2006.
3 Qinfeng Shi, Yanning Zhang, Linear Feature Detection based on Beamlet Analysis, in Proc. of the Third National Conference on Signal and Information Processing (Chinese), 2004, pp. 206-209
2 Qinfeng Shi, Ying Li, Yanning Zhang, A New Automatic Segmentation for Synthetic Aperture Radar Images, in Proc. of the International Symposium on Intelligent Multimedia, Video & Speech Processing (ISIMP'04), 2004, pp. 739-742
1 Qinfeng Shi, Yanning Zhang, Adaptive Linear Feature Detection based on Beamlet, in Proc. of the Third International Conference on Machine Learning and Cybernetics (ICMLC'04), 2004, pp. 3981-3984.

Talks

  1. Discriminative Human Action Segmentation and Recognition using Semi-Markov Model, NEC lab in Princeton and CSML in University College London, July 2008
  2. Semi-Markov Model for Sequential Data Analysis, Computer Science Lab, HKUST, July 2007
  3. Introduction to Conditional Random Fields, Computer Science Lab, HKUST July 2007
  4. Automatic Paragraph Segmentation via Semi-Markov Models, EMNLP 07, Prauge, Jun. 2007 [pdf]
  5. Introduction to Cover Tree, SML, Canberra, Oct. 2006 [pdf]

Teaching Assistant

  1. Introduction to Machine Learning COMP4670/6467, Australian National University, Semester 1, 2007

  2. Network Information System COMP2410/6340, Australian National University, Semester 1, 2007

Software

  1. Hash Kernels. Check out the latest version of the source code with this command:

    svn co http://elefant.developer.nicta.com.au/local/repos/trunk/elefant/stream
  2. Semi-Markov Models for Sequence Segmentation and Classification (reconstructing ... Download the source code and synthetic data. )

    Semi-Markov Models package can be used for sequence segmentation and classification. As applications, it has been applied to Automatic Paragraph Segmentation and Human Action Segmentation & Recognition. It is implemented in C++ based on SVM-Struct.

People

Links