Junren Chen
 |
PhD Student |
Profile
Welcome! I am Junren Chen (Chinese: 陈军任), a final-year PhD student at Department of Mathematics, The University of Hong Kong. I am very fortunate to have Prof. Michael K. Ng as my advisor. I was also fortunately a visiting graduate student at NUS with Prof. Jonathan Scarlett, at Columbia University with Prof. Ming Yuan. Expected to graduate in 2025 June, I am looking for postdoc position starting Fall 2025.
Research Interests
My research interests include high-dimensional statistics, nonconvex optimization, mathematics of data science, signal processing and machine learning. I study both the information-theoretic and the algorithmic aspects of a specific problem. My Ph.D. research has been focused on estimation under nonlinearities, most typically, quantization. I would be thrilled to take steps towards optimization, applied probability, deep learning theory and so on.
Selective Papers
Below are some representative papers with links and possibly short notes. See Google Scholar for full list.
High Dimensional Statistical Estimation under Uniformly Dithered One-Bit Quantization. [T-IT] [Arxiv]
J. Chen, C.-L. Wang, M. K. Ng, D. Wang.
IEEE Transactions on Information Theory, 2023.
[Propose to incorporate truncation before the quantization of heavy-tailed data in high-dimensional estimation]
Exact Thresholds for Noisy Non-Adaptive Group Testing. [Arxiv]
J. Chen, J. Scarlett.
ACM-SIAM Symposium on Discrete Algorithms (SODA 25).
[Exact thresholds for both Bernoulli design and near-constant weight design; noisy counterpart for this SODA paper]
A Parameter-Free Two-Bit Covariance Estimator with Improved Operator Norm Error Rate. [Arxiv]
J. Chen, M. K. Ng.
In Revision
[The first two-bit covariance estimator with dimension-free error rate for sub-Gaussian distributions; this AoS paper opens this direction]
One-Bit Phase Retrieval: Optimal Rates and Efficient Algorithms. [Arxiv]
J. Chen, M. Yuan.
In Review
[Near optimal results for (sparse) signal recovery from the phaseless bits \(\mathbf{y}=\textrm{sign}(|\mathbf{Ax}|-\tau)\); phaseless counterparts of the memoryless 1-bit compressed sensing theory]
Optimal Quantized Compressed Sensing via Projected Gradient Descent. [Arxiv]
J. Chen, M. Yuan.
In Review
[Projected Gradient Descent is in general optimal for quantized compressed sensing; extend the optimality of NBIHT in this JACM paper to general models \(\mathbf{y}=\mathcal{Q}(\mathbf{Ax})\) and general signal structures]
Phase Retrieval of Quaternion Signal via Wirtinger Flow. [T-SP]
J. Chen, M. K. Ng.
IEEE Transactions on Signal Processing, 2023.
[Among the first studies of phase retrieval of (pure) quaternion signals]
A Unified Framework for Uniform Signal Recovery in Nonlinear Generative Compressed Sensing. [NeurIPS]
J. Chen, J. Scarlett, M. K. Ng, Z. Liu.
Neural Information Processing Systems (NeurIPS), 2023.
[Generative priors counterpart of this FoCM paper, with sharper rates]