Dehua Cheng

Dehua Cheng

Ph.D. Candidate

Computer Science Department
Viterbi School of Engineering
University of Southern California

Email: dehuache_AT_usc.edu
Office: PHE 316

My name is Dehua Cheng (程德华). I am a PhD candidate in the Computer Science Department, of University of Southern California. I work on Machine Learning under supervision of Prof. Yan Liu.

My primary research interest lies in large scale machine learning. More specifically, I am interested in developing linear or sublinear numerical routines for machine learning algorithms by exploiting randomization and the structures of both the machine learning model and data. I have also worked on parallel inference for probabilistic graphical model, including topic models, Bayesian nonparametrics, etc. And in general, I am interested on (1) designing computational efficient models, and (2) improving existing machine learning algorithms from the computational efficiency aspect. My CV can be found here.

Experience

May 2017 - Aug. 2017

Software Engineer Intern
Feed Machine Learning @ Facebook
Advisor: Qichao Que

May 2016 - Aug. 2016

Research Intern
Thomas J Watson Research Center
IBM Research
Advisor: Jie Chen

Education

Aug. 2012 - present

Ph.D. Candidate
Computer Science Department
University of Southern California
Advisor: Yan Liu

Aug. 2008 - Jul. 2012

B.S.
Mathematics and Physics
Tsinghua University
Thesis advisor: Changshui Zhang

Publications

Preprints

Michael Tsang, Dehua Cheng, and Yan Liu, Detecting Statistical Interactions from Neural Network Weights, NIPS IEVDL 2017, and arXiv:1705.04977

Dehua Cheng*, Natali Ruchansky*, and Yan Liu (*Equal Contributions), Matrix Completion with Graphs: Identifying Completable Submatrices via Edge Connectivity.

Jie Chen, Dehua Cheng, Yan Liu, On Bochner's and Polya's Characterizations of Positive-Definite Kernels and the Respective Random Feature Maps, Submitted to JMLR. arXiv:1610.08861

Dehua Cheng, Yu Cheng, Yan Liu, Richard Peng, and Shang-Hua Teng, Spectral Sparsification of Random-Walk Matrix Polynomials, arXiv:1502.03496.

Dehua Cheng, Yu Cheng, Yan Liu, Richard Peng, and Shang-Hua Teng, Scalable Parallel Factorizations of SDD Matrices and Efficient Sampling for Gaussian Graphical Models, arXiv:1410.5392.

Refereed Publications

Dehua Cheng, Richard Peng, Ioakeim Perros, and Yan Liu, SPALS: Fast Alternating Least Squares via Implicit Leverage Scores Sampling, NIPS 2016. [PDF] [Code]

Dehua Cheng, Yu Cheng, Yan Liu, Richard Peng, and Shang-Hua Teng, Efficient Sampling for Gaussian Graphical Models via Spectral Sparsification, COLT 2015. [PDF].

Qi Yu, Dehua Cheng, and Yan Liu, Accelerated Online Low Rank Tensor Learning for Multivariate Spatiotemporal Streams, ICML 2015. [PDF]

Dehua Cheng*, Xinran He*, and Yan Liu (*Equal Contributions), Model Selection for Topic Models via Spectral Decomposition, AISTATS 2015. [PDF]

Dehua Cheng, Mohammad Taha Bahadori, and Yan Liu, FBLG: A Simple and Effective Approach for Temporal Dependence Discovery from Time Series Data, KDD 2014. [PDF]

Dehua Cheng, and Yan Liu, Parallel Gibbs Sampling for Hierarchical Dirichlet Processes via Gamma Processes Equivalence, KDD 2014. [PDF]

Services

Student Volunteer, NIPS ’16

Workshop organizer, MiLeTs @ KDD ’16

Student Volunteer, ICML ’15

Student Volunteer, KDD ’14