Liu, Ji (刘霁)

Tel: 480-388-9026

Email: ji-liu@cs.wisc.edu, (Wisconsin); ji.liu@asu.edu (ASU)

I am currently the third year student supervised by Stephen J. Wright in the Department of Computer Sciences of University Wisconsin-Madison. Before joining UW, I learned from Jieping Ye and Peter Wonka and obtained my master degree majoring in Computer Science at Arizona State University. My current research interests include machine learning, optimization, and computer vision. 

Research Interests

    Machine Learning (sparse learning: theory and algorithm)

    Optimization (asynchronous methods, large scale problem, efficient first order algorithm, stochastic algorithm)

    Computer Vision (tensor completion, image inpainting, image segmentation, stereo vision)

Education

University of Wisconsin, Madison        09/2010-present

           PHD candidate, in Computer Sciences

Arizona State University (ASU)       09/2008-08/2010

           Master Degree in Computer Science         

           Advisors: Jieping Ye and Peter Wonka                                        

University of Science and Technology of China (USTC)       09/2001-07/2005

           Bachelor degree in Automation                                                                                                

           Minor degree in Business Management                                                   

Publications

(09/2008-present, USA):

21. Ji Liu, Yuan Lei, and Jieping Ye,  “Guaranteed Sparse Recovery under Linear Transformation”, ICML, 2013.

20. Yang Cong, Ji Liu, Junsong Yuan, and Jiebo Luo,  “Self-supervised Online Metric Learning with Low Rank Constraint for Scene Categorization”, IEEE Transaction on Image Processing, 2013.(To appear)

19. Bo Liu, Sridhar Mahadevan, and Ji Liu,  “Regularized Off-Policy TD-Learning”, NIPS, 2012.(Spotlight)

18. Ji Liu and Stephen J. Wright, “Robust Dequantized Compressive Sensing”, Arxiv preprint arXiv:1207.0577, 2012.

17. Jianhui Chen, Ji Liu, and Jieping Ye, “Learning Incoherent Sparse and Low-Rank Patterns from Multiple Tasks”, ACM Transaction on Knowledge Discovery from Data, 2012.

16. Ji Liu, Peter Wonka, and Jieping Ye, “A Multi-stage Framework for Dantzig Selector and Lasso”, Journal of Machine Learning Research, 2012.

15. Yang Cong, Junsong Yuan, Ji Liu, "Abnormal Event Detection in Crowded Scenes Using Sparse Representation", Pattern Recognition, 2012. (To appear)

14. Ji Liu, Przemyslaw Musialski, Peter Wonka, and Jieping Ye, “Tensor Completion for Estimating Missing Values in Visual Data”, IEEE Transaction on Pattern Analysis and Machine Intelligence, 2012. [code]

13. Yang Cong, Junsong Yuan, Ji Liu, "Sparse Reconstruction Cost for Abnormal Event Detection", CVPR, 2011.

12. Ji Liu, Peter Wonka, and Jieping Ye, “Multi-stage Dantzig Selector”, NIPS, 2010. [code]

11.  Jianhui Chen, Ji Liu, and Jieping Ye, "Learning Incoherent Sparse and Low-Rank Patterns from Multiple Tasks", KDD, 2010. (Honorable Mention for the best research paper)

10. Ji Liu, Przemyslaw Musialski, Peter Wonka, and Jieping Ye, “Tensor Completion for Estimating Missing Values in Visual Data”, ICCV, 2009. [video] [code]

9.   Ji Liu, Jun Liu, Peter Wonka, and Jieping Ye, “Sparse Non-negative Tensor Factorization Using Columnwise Coordinate Decent”, 45(1), 649-656, Pattern Recognition, 2012. [code]

(03/2007-09/2008, China):

8.    Ji Liu, Yang Cong, Yuechao Wang, Yandong Tang, “Lunar Terrain Reconstruction Based on PDEs Method”, 2008 IEEE International Conference on Image Processing (ICIP), pp1504-1507. San Diego, America, Oct. 2008.

7.   Ji Liu, Junjian Peng, Yuechao Wang, Yandong Tang, “A PDEs Method Preserving Boundaries on Dense Disparity Map Reconstruction”, 3^rd Conference of Computer Vision Theory and Application (VISAPP). Madeira, Portugal, pp655-661, Jan. 2008.

6.   Ji Liu, Yuechao Wang, Chuan Zhou, Yanfeng Geng, “A Navigation Simulation System of Lunar Rover”, 2008 IEEE International Conference on Networking, Sensing and Control (ICNSC), pp556-561, Sanya, China, Apr. 2008. [video1] [video2]

5.    Ji Liu, Yuechao Wang, Chuan Zhou, Yongzhi He, “A Method of Eliminating the Wheel-terrain Interaction Errors in Lunar Rover Simulation”, Chinese Journal of System Simulation, ISSN 1004-731X, vol. 14, 2008.

4.   Yang Cong, Xiaomao Li, Ji Liu, YandongTang, “A Stairway Detection Algorithm based on Vision for UGV Stair Climbing”, 2008 IEEE International Conference on Networking, Sensing and Control (ICNSC), pp1806-1811, Sanya, China, Apr. 2008.

3.    Jujian Peng, Ji Liu, Yanfeng Geng, Jianda Han, Yandong Tang, “A Dynamic Stereo Matching Method Based on Epipaolar-region”, Chinese Journal of Computer engineering, ISSN 1000-3428, vol.24, 2008.

2.   Yanfeng Geng, Kai Kang, Ji Liu, and Hong Wang, “Manufacturing Schedule of Dual-armed Cluster Tools Based on Heuristic Search”, International Conference on Industry Technology (ICIT), Chengdu, China, Apr. 2008.

1.   Yongzhi He, Chuan Zhou, Ji Liu, Dalong Tan, “Research on Movement Simulation for Wheeled Mobile Robot”, Chinese Journal of Scientific Instrument, 4(28), pp217-220, 2007.

Research & Project Experiences                                                                                         

(06/2012-09/2012 at NEC Media Analytics Department, USA):
        Machine Learning and Optimization

        Greedy methods in feature selection and its theoretical analysis
　

(10/2010-present at UW-Madison, USA):
        Machine Learning and Optimization

        Accelerated randomized Kaczmarz algorithm

        Robust dequantized compressive sensing

        Stochastic optimization
      Computer Vision

        Abnormal event detection
　

(09/2008-08/2010 at ASU, USA):

Sparse Learning                                                                                                     

  Designed a multi-stage algorithm based on the standard Dantzig selector

  Theoretically proved it outperforms the standard Dantzig selector

Tensor Completion                                                                                            

Proposed a definition for the trace norm of tensor

Designed an algorithm to estimate the missing value in tensor data by minimizing the trace norm of the tensor

Applied the proposed algorithm on image inpainting and biological data completion

Dimensionality Reduction                                                                                       

Proposed a fast (5-10times faster than the previous algorithms) sparse non-negative tensor factorization algorithm using columnwise coordinate decent (CCD).

Applied the proposed algorithm to extract the features in high dimensional data such as the embryo image data set

(09/2006-04/2008 at Chinese Academy of Science, China):

Stereo Vision

Proposed a general functional model including the prior information from image segmentation, in order to preserve the discontinuity on the boundaries.

Applied the proposed model to reconstruct the lunar terrain and register the medical images

Lunar Rover Simulation and Virtual Reality [video1] [video2]

Build up the software platform for the lunar rover navigation simulation system

Applied the numerical optimization technology to fix the error of the traditional kinematics model in system simulation (this intractable problem had puzzled our group for three months before my participation).

Reconfigurable Multi-modules Robot [video1] [video2]           

Designed the multi-module coordinative motion strategy for the robot

Implemented most control algorithms in the hardware platform

3D Vessel Segmentation [video1] [video2]                     

Extended the “snack” model to 3 dimensional case to segment the vessel in 3D data

Honors & Awards

Honorable mention of the best research paper in KDD, 2010

NIPS Travel Grant, 2010

PAMI TC Travel Grant, 2009

Grade 3, National Graduate Mathematical Contest in Modeling, 2007

First Prize, Mathematics Contest in Modeling of Northeast China, 2007 (7/2000+)

Excellent Student Scholarship, CAS, 2007

Graduate Fellowship with Honor, CAS, 2006

Outstanding Students Scholarship, USTC, 2003, 2004

Activities

Vice-Chairman of graduate Union in SIA, CAS, 2006-2008  

Practice Ministry of Student Union, USTC, 2004-2005

Vice-Chairman of Student Union of Department of Automation, USTC, 2002-2003

Chairman of Student Union, No. 2 Middle School of Wanzhou, 1999-2000

Skills

Matlab, C, C++, Python, OpenGL, Java, Windows, Linux

Mathematical courses taken

           Linear Algebra

           Advanced Mathematics 1 (Calculus)

           Advanced Mathematics 2 (Calculus)

           Function of Complex Variable

           Probability Theory and Mathematical Statistics

           Mathematics Physics Equations

           Stochastic Processes

           Optimization Methods and Applications in Engineering

           Optimum Control

           Stochastic Estimation and Control

           Differential Equation and Image Processing

           Algebra Theory in Control Science

           Real Function Theory & Functional Analysis

           Mathematical Theory in Control Science (Information Geometry)

           Optimization Algorithms w/Engineering Applications

           Bayesian Data Analysis

           Convexity Analysis

           Nonlinear Optimization I

           Nonlinear Optimization II

           Variation Analysis

           Stochastic Optimization