Renfeng Peng
Renfeng Peng is a Research Assistant at the Department of Data Science, City University of Hong Kong (CityUHK), working with Prof. Fenglei Fan. He completed his Ph.D. in 2026 at the Academy of Mathematics and Systems Science (AMSS), Chinese Academy of Sciences (CAS), under the supervision of Prof. Ya-xiang Yuan. He received his bachelor’s degree in mathematics and applied mathematics from the School of Mathematical Sciences, Tongji University, in 2021. His research interests include tensor optimization, manifold optimization, and machine learning.
Updates
- new preprint “Dynamics as Code: On Model Compression via Dynamic System.”
- I am happy to join Prof. Fenglei Fan's group at City University of Hong Kong as a research assistant.
- I have successfully defended my PhD dissertation.
Earlier updates
- new preprint “Quotient geometry of tensor ring decomposition.”
- new preprint “Normalized tensor train decomposition.”
- new paper accepted in SIAM Journal on Matrix Analysis and Applications - “Optimization on product manifolds under a preconditioned metric.”
- new conference paper accepted in ISIT 2025- “Riemannian optimization for Holevo capacity.”
- new preprint - “Riemannian optimization for Holevo capacity.”
Education
-
2021.09 – 2026.07
University of Chinese Academy of Sciences
Academy of Mathematics and Systems Science
-
2017.09 – 2021.07
Tongji University
School of Mathematical Sciences
Refereed papers & preprints
- Dynamics as Code: On Model Compression via Dynamic SystemarXiv preprint arXiv:2610.11115, 2026
- High-dimensional extreme eigenvalue problems: low-rank tensor parametrization and optimizationarXiv preprint arXiv:2609.30138, 2026
- A second order numerical scheme for fractional option pricing modelsEast Asian J. Appl. Math, 2021
LRTCTR
A Matlab solver for low-rank tensor completion in tensor ring decomposition.popman
A Matlab solver for preconditioned Riemannian optimization methods on product manifolds.TRAM
Tucker rank-adaptive method for Tucker tensor completion.TuckerDesing
A Matlab solver for low-rank Tucker tensor optimization via desingularization.TuckerApoFree
First-order methods for low-rank Tucker tensor optimization converging to stationary points.GeomNTT
A Matlab solver for optimization on NTT tensors.Selected presentations
| Date | Place | Conference |
|---|---|---|
|
2026/04
|
Shanghai, China
|
2026 International Conference on Optimization: Techniques and Applications |
|
2025/08
|
Harbin, China
|
The 15th International Conference on Numerical Optimization and Numerical Linear Algebra |
|
2025/07
|
Los Angeles, U.S.
|
International conference on continuous optimization |
|
2024/11
|
Beijing, China
|
The 7th PKU PhD Symposium for Applied Mathematics |