Yongxin Taylor Xi

I am a fourth year Ph.D. candidate in the ISS division of Department of Electrical Engineering at Princeton University. My advisor is Professor Peter J. Ramadge. My research focuses on developping useful and theoritically justifiable algorithms in machine learning, over a broad spectrum of unsupervised, supervised and semi-supervised learning. Other areas such as optimization theory, probability theory, linear algebra and differential geometry generate the main sources of my algorithmic ideas; while image processing, computer vision and cognitive neuroscience are the main applications of these algorithms.

What is the limit of artificial intelligence? Can a computer/machine learn as well as human do, or even better and faster? We do not know yet. As a researcher working enthusiastically to push further the performance boundary of various machine learning algorithms, I have not yet been threatened by the day to come. But I do know that machine learning as a tool is helping the world tremendously almost everywhere. A supervised learning algorithm can be used to detect patterns in the process of learning and conduct inference by itself later on. That is what face detection/recognition, computer-aided medical diagnoses, financial market prediction are all about. On the other hand, unsupervised learning such as clustering helps scientists and engineers to better organize and interpret the experimental data to make meaningful inferrence and new discoveries.

I like collaborating with people from different places. People have different ideas and perspectives. It is enjoyable to be motivated and inspired by them from time to time. The person I collaborate most, i.e. my advisor Prof. Ramadge, is an immediate justification for this. Besides him, I have collaborated with Prof. Robert Schapire in the computer science department on a sparse boosting paper. I have also worked together with several group members on different papers. Now I am collaborating with Dr. Ray Lee in the psychology department on developping online SVM algorithms for real-time fMRI analysis of the interative thinking of human brains.

Besides me, the Ramadge research group has other 6 graduate students: Eugene Brevdo, Bryan Conroy, Zhen James Xiang, Axelander Lorbert, Hao Xu and David Eis. We have a regular group meeting every Friday noon at Equad F216 to discuss most recent researches on the common basis.

One more thing: if you can read Chinese and like novels, check out my recent novel "Left Or Right" (Click here to see the novel)!

Publications

2010
LOW RANK PROBABILISTIC CLUSTERING (LoRaC)
Y. T. Xi, P. Ramadge
Ready to be submitted to the Thirteenth International Conference on Artificial Intelligence and Statistics (AISTATS 2010), May 13-15, Sardinia, Italy

A SUPERVISORY APPROACH TO SEMI-SUPERVISED CLUSTERING
B. Conroy, Y. T. Xi, P. Ramadge
Submitted to IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2010), March 14-19, Dallas, Texas, USA

2009
USING SPARSE REGRESSION TO LEARN EFFECTIVE PROJECTIONS FOR FACE RECOGNITION(.pdf)
Y. T. Xi, P. Ramadge
IEEE International Conference on Image Processing (ICIP 2009), November 7-10, Cairo, Egypt

BOOSTING WITH SPATIAL REGULARIZATION (.pdf)
Z. J. Xiang, Y. T. Xi, U. Hasson, P. J. Ramadge
Advances in Neural Information Processing Systems (NIPS 2009), December 7-10, Vancouver, B.C., Canada

SPEED AND SPARSITY OF REGULARIZED BOOSTING (.pdf)
Y. T. Xi, Z. J. Xiang, P. Ramadge, R. Schapire
Twelfth International Conference on Artificial Intelligence and Statistics (AISTATS 2009), April 16-19, Clearwater, FL, USA

SEPARABLE PCA FOR IMAGE CLASSIFICATION (.pdf)
Y. T. Xi, P. Ramadge
IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2009), April 18-24, Taipei, Taiwan

2008
Y. T. Xi, Z. J. Xiang, P. Ramadge, R. Schapire, Poster Abstract: Speed and Sparsity of Regularized Boosting (.pdf)
Woman in Machine Learning Workshop (WIML 2008), December 8, Vancouver, B.C., Canada

Y. T. Xi, P. Ramadge, Poster: Using LARS for learning effective projections in face recognition (.pdf)
Computer Research Association Workshop (CRA-W 2008), March 13-14, Seattle, WA, USA

Coursework

Here is the list of courses that I have taken so far, * standing for my favorite ones .

Fall 20009:

ORF551**: Probability Theory (Instructor: Erhan Cinlar)
ELE301: Digital Signal Processing (as TA; Instructor: Bede Liu)

Spring 2009:

COS513**: Foundations of Graphical Models (Instructor: David Blei)
COS598D**: Advanced Topics: Boosting: Foundations and Algorithms

Fall 2008:

MAT314*: Real Analysis (Instructor: Dmitri Beliaev)

Spring 2008:

COS511**: Theoretical Machine Learning (Instructor: Robert Schapire)
WRI501*: Princeton Writing Program (Instructor: Xenia Morin)

Fall 2007:

COS429*: Computer Vision (Instructor: Fei-Fei Li)
MAT327*: Introduction to Differential Geometry (Instructor: Anna Wienhard)
ELE488*: Image Processing (as TA, won Outstanding TA Award; Instructor: Bede Liu)
SEAS shortcourse: Programming in MATLAB (as TA; Instructor: Peter Ramadge, K. Sircar, Kevin Wayne)

Spring 2007:

ELE571: Digital Neurocomputing (Instructor: Sun-Yuan Kung)
ELE539A: Special Topics in Informations Sciences and Systems: Optimization of Communication Systems (Instructor: Mung Chiang)
PSY259B*: Introduction to Cognitive Neuroscience (Instructor: Sabine Kastner)
ELE530**: Theory of Detection and Estimation (Instructor: Peter Ramadge)

Fall 2006:

ELE521*: Linear System Theory (Instructor: Peter Ramadge)
ORF309**: Probability and Stochastic Systems (Instructor: Erhan Cinlar)
ELE525: Random Processes in Information Systems (Instructor: Hisashi Kobayashi)
ELE465: Switching and Sequential Systems (Instructor: Sun-Yuan Kung)

Presentations

Poster for AISTATS 2009

Speed and Sparsity of Regularized Boosting (.pdf)

Poster for ICASSP 2009

Separable PCA for Image Classification (.pdf)

NIAM Talk, Psychology Dept., Princeton university

April 25, 2008
----FMRI Classification: Smart Use of LASSO to win over Ridge Regression(.pdf)

CRA-W Workshop Poster Presentation, Seattle, WA:

March 13, 2008
----Using Lars for learning effective projections in face recognition(.pdf)

1st PIXL Retreat Presentation, Grouds for sculptures, NJ:

Sep 13, 2007
----Support Vector Decomposition Machine(.pdf)

Oral General Exam Presentation, B327, Equad, Princeton

Jun 5, 2008
----Sparse Learning from Compressible Data Sets

Ramadge Group Presentations:

Oct 16, 2009
----Low Rank Probabilistic Clustering (slides available upon acceptance of the paper)

July 24, 2009
----Matrix Completion(.pdf)

October 10, 2008
Separable Principle Component Analysis (.pdf)

August 15, 2008
New 2D methods in pattern analysis (.pdf)

July 11, 2008
Speed and Sparsity of Regularized Boosting(.pdf)

June 19, 2008
Paritial slides occur in: Advanced Signal Processing for Functional MRI Analysis, by Prof. Ramadge at University of Newcastle (.pdf)

Mar 7, 2008
----Using Lars for learning effective projections in face recognition (.pdf)

The following two presentations are closely related with the paper "The Geometry of Algorithms with Orthogonality Constraints," by A. Edelman, T. Arias, and S.T. Smith (.pdf)

Nov 2, 2007
----Optimization on Manifold (.pdf)

Nov 16, 2007
----Detailed study of Stiefel Manifold and associated Gradient Descent optimization (notes.pdf)

Jul 12, 2007
----Singular Value Decomposition Machine with various problem formulations(.pdf)

May 22, 2007
----Learning from fMRI data via Support Vector Decomposition Machine(.pdf)

Mar 24, 2007
----The Ventral Visual Object Pathway in humans: Evidence from fMRI (.pdf)

Projects

COS513 Term Project: Semi-supervised Clustering Using Affinity Propagation (.pdf)
COS511 Term Project: New Results on L1 Regularized Boosting (.pdf)
COS429 Term Project: Discriminative Face Recognition Via Regression (.pdf)
MAT327 Term Project: A novel method to derive Geodesics and Parallel Transport on Stiefel Manifold (part I:.pdf,part II:.pdf)
ELE539 Project Paper: Alternative Ways to Solve Optimization Problem in Support Vector Decomposition Machine (Midterm report(.pdf), Final report(.pdf))
ELE 571 Term Project: A comparison study of Model based and Model free Linear Cancer classifiers (.pdf)
Singular Value Decomposition Machine Theory Study (.pdf)

Curriculum Vitae

Socialwork

Social Chair, GCHC (Graduate College House Committee).
Nov 2007--Oct 2007.
During my term, I held the weekly social hour (cocktail section) in Old Graduate College, and maintained a House Film Library of video cassettes and digital media for resident members.

Vice President, ACSSPU (Association of Chinese Students and Scholars).
Sep 2007--Sep 2008.
ACSS is a student organization at the Princeton University representing ethnic Chinese students and scholars within the Princeton community. Our members include over 300 students, visiting scholars, faculty and staff on campus, and scores of friends of the association living in the Princeton area. We organize social, intellectual and cultural activities regularly among Chinese students and scholars at Princeton University as well as other interested members in the Princeton Community. I was the director in general for the Mid Autumn Gala Show 2007. Some photo shots can be found here.

President, CIEPU (Chinese Institute of Engineers at Princeton University).
Dec 2007--now.

Vice President, Graduate Engineering Ambassador, Princeton University (Graduate Engineering Ambassadors).
Jan 2009--now.

Special Events Chair, Graduate College House Committee.
March 2008--March 2009.
Organize special events for graduate students, such as the Formal, summer parties, Halloween party, and the Swing Night.

President, DCSPU (Distinguished Citizen Society at Princeton University).
Mar 2009--now.

Mini Biography

I came to the United States for the first time at the end of August 2006. Before that I had been raised from the beautiful city Shanghai in China. I did both middle school and high school in Shanghai No.3 Girls' Middle School, which was set up with the name St. Maria Girls' Middle School by an American protestant Christian church in 1892. Then I got a Bachelor degree in both Electrical Engineering and Computer Science in Fudan University, Shanghai.

Since I came to Princeton, I fell in love with this lovely town, the university, and all the people around. I believe I've made a good decision, and I am enjoying the life here each day both academically and socially. It is hard to predict where I might go after the PhD degree, but I know I am heading toward my dream.

My main hobbies are traveling, pet raising, sports (tennis, ice skating) and Chinese calligraphy. So far I have been to Yellowstone NP and Grand Teton NP (Wyoming, Montana, Idaho, Utah), Hawaii Islands (Oahu and Kauai), Niagara Falls in New York, San Francisco in California (for CRA-W workshop), New Milford in Connecticut (where my cousin and her family lives), Las Vegas in Nevada, Houston in TX, Orlando in Florida and Disney World (Grace Hopper Conference), Acadia National Park in Maine (outing with friends), Boston in Massachusetts (visiting friends), Kansas City in Missouri (visiting an elder friend), Washington D.C. (3 day trip with my parents), Seattle in Washington (CRA-W), New York City and Philadelphia of course. I am going to Vancouver, Canada for conference soon! I also like to make short Video Clips for fun, to entertain and to collect shiny pieces of lives.

I have had lots of pets since I was young. I just cannot help myself with their cuteness. I have a guinea pig at home. His name is Coffee, one and a half year old now. Oh I love him so much. Recently I had two new birdies: Kiki and Wiwi. They are zebra finches. Although I am no real expert on any sports, tennis, ice-skating, ballroom dancing, squash, table tennis and swimming are my favorite ones. Try playing tennis till soaked out, and try ice-skating on the frozen Carnegie Lake!