F310 E-Quad, Olden Street,
Princeton, NJ 08544 USA
Machine Learning, Scattering Transforms, Sparse Representation, Deep Neural Networks
erieure, Paris, France
Oct. 2013 - Present
Visiting Ph.D. Student in Computer Science. Advisor: Prof. St´ephane Mallat.
Princeton University, NJ, United States
May. 2012 - Present
Ph.D. Candidate in Electrical Engineering. Advisor: Prof. Peter J. Ramadge.
Princeton University, NJ, United States
M.A. in Electrical Engineering, GPA: 3.92/4.00.
Sep. 2010 - May. 2012
Hong Kong University of Science and Technology, HK Aug. 2008 - Dec. 2008
Exchange Student in Electrical and Computer Engineering, GPA: 4.00/4.00.
Tsinghua University, Beijing, China
Sep. 2006 - Jul. 2010
B.E. in Electronic Engineering, GPA: 92.1/100, ranked 5th/306.
May. 2012 - Oct. 2012
Alcatel-Lucent Bell Labs, Stuttgart, Germany
• Formulated the prediction of a user’s next cell in cellular networks as a classification
problem based on Channel State Information (CSI) and handover history.
• Introduced a new machine learning based prediction system to anticipate the cell a
user will hand-over to.
• Obtained at early prediction time substantially higher prediction accuracy than
previous state-of-the-art methods.
Assistant in Instruction
Sep. 2011 - Jan. 2012
Department of Computer Science, Princeton University
• COS402: Artificial Intelligence.
• Instructor: Prof. Robert Schapire.
Department of Computer Science, Ecole
Learning with Deep Scattering Networks
Nov. 2013 - Sep. 2014
• Reintroduced scattering transforms as a product of multirate filter bank operators.
• Studied orthogonal Haar scattering with multiple trees and its contraction properties.
Introduced oversampled Haar scattering and explored its relationship with continuous
• Introduced a learnt deep scattering network with Haar wavelets, which computes
invariants with iterated contractions adapted to training data. It therefore defines
a deep convolution network model, whose contraction properties can be analyzed
Department of Electrical Engineering, Princeton University
Sparse Representation and Classification
Oct. 2012 - Nov. 2013
• Studied the key attributes that lead to the success of sparse representation based
classification by digging into the essential differences among several dictionary based
• Proposed an effective music genre classication approach based on sparse coding and
scattering representations, achieving one of the best classification results so far on
the public accessible dataset GTZAN.
• Designed a new weighted sparse representation-based voting scheme that yields
accuracy improvements for music genre classification.
Sequential Lasso Screening
Mar. 2013 - Jul. 2013
• Collaborator: Yun Wang
• Designed a feedback-controlled iterative dictionary screening scheme that speeds up
solving the lasso problems more robustly.
• Proposed an early decision mechanism for classification tasks, based on sequential
lasso screening and sparse representation-based classfier, which for several decision
tasks yields improved classication accuracy and considerable computational speedup.
Low-Power Learning Algorithms
Mar. 2011 - Jul. 2011
• Collaborator: Pingmei Xu, Sun-Yuan Kung
• Designed a hierarchical classification scheme that used lower-order kernel-based
classifiers to deliver performance comparable to higher-order methods but reduced
the computation complexity by ×101∼3 depending on specific applications.
Xu Chen, Xiuyuan Cheng and St´ephane Mallat, Unsupervised Deep Haar Scattering
on Graphs, Conference on Neural Information Processing Systems (NIPS’14), Dec 8th
- 11th, 2014, Montreal, Quebec, Canada
Xu Chen and Peter J. Ramadge, Collaborative Representation, Sparsity or
Nonlinearity: What is Key to Dictionary Based Classification?, IEEE
International Conference on Acoustics, Speech and Signal Processing (ICASSP’14), May
4th - 9th, 2014, Florence, Italy
Yun Wang, Xu Chen and Peter J. Ramadge, Speeding Up Sparse Representation
Classication via Sequential Screening, 1st IEEE Global Conference on Signal and
Information Processing (GlobalSIP’13), Dec. 3rd - 5th, 2013, Austin, TX, United States
Xu Chen, Fran¸cois M´eriaux and Stefan Valentin. Predicting a User’s Next Cell
with Supervised Learning Based on Channel States, 14th IEEE International
Workshop on Signal Processing Advances in Wireless Communications (SPAWC’13),
Jun. 16th - 19th, 2013, Darmstadt, Germany
Xu Chen and Peter J. Ramadge, Music Genre Classification Using Multiscale
Scattering and Sparse Representation, 47th Annual Conference on Information
Sciences and Systems (CISS’13), Mar. 20th - 22nd, 2013, Baltimore, MD, United States.
Xu Chen, Xiuyuan Cheng and St´ephane Mallat, Deep Networks with Adapted
Haar Scattering, UCL-Duke Workshop on Sensing and Analysis of High-Dimensional
Data (SAHD’14), Sep 4th - 5th, 2014, London, UK
Chateaubriand Fellowship, Embassy of France, Washington, D.C., US
RISE professional Scholarship, DAAD, Germany
Graduate Fellowship, Princeton University, NJ, US
Suzhou Industry Park Scholarship, Tsinghua University, China
Zhou Huiqi Scholarship, Tsinghua University, China
National Scholarship, Ministry of Education, China
Languages: Chinese (Native), English (Fluent), French (Elementary)
Programming: Matlab, C/C++, Java.
Operating Systems: Linux/Unix, Mac OS X, Windows.