Yijun Dong (董一珺)

Courant Instructor/Assistant Professor (postdoc) Courant Institute of Mathematical Sciences New York University Email: yd1319 [@] nyu [DOT] edu Office: WWH 526, 251 Mercer St, New York, NY 10012 (Curriculum Vitae, Google Scholar, GitHub)

About Me

I am a Courant Instructor/Assistant Professor (postdoc) at the Courant Institute of New York University. I completed my PhD at the Oden Institute of UT Austin, advised by Prof. Per-Gunnar Martinsson and Prof. Rachel Ward.

My research lies in randomized numerical linear algebra and learning theory. I am broadly interested in high-dimensional problems with low intrinsic dimensions, with focuses on the computational and sample efficiency of algorithms in machine learning and scientific computing. For computational efficiency, my work is centered on randomized algorithms for dimension reduction and low-rank approximation. For sample efficiency, my work focuses on the generalization and distributional robustness of learning algorithms in data-limited settings.

News

• 2025/02: New preprint “Discrepancies are Virtue: Weak-to-Strong Generalization through Lens of Intrinsic Dimension” on arXiv

• 2025/02: “Greedy Output Approximation: Towards Efficient Structured Pruning for LLMs Without Retraining” accepted at CPAL 2025

• 2024/09: “Sketchy Moment Matching: Toward Fast and Provable Data Selection for Finetuning” accepted at NeurIPS 2024

• 2024/06: “Efficient Bounds and Estimates for Canonical Angles in Randomized Subspace Approximations” accepted at SIAM Journal on Matrix Analysis and Applications

Selected Publications

(* denotes equal contribution or alphabetical order)

• Discrepancies are Virtue: Weak-to-Strong Generalization through Lens of Intrinsic Dimension
  Yijun Dong, Yicheng Li, Yunai Li, Jason D. Lee, Qi Lei
  Preprint, 2025.
  

• Sketchy Moment Matching: Toward Fast and Provable Data Selection for Finetuning
  Yijun Dong*, Hoang Phan*, Xiang Pan*, Qi Lei
  Neural Information Processing Systems (NeurIPS), 2024. [GitHub]
  

• Robust Blockwise Random Pivoting: Fast and Accurate Adaptive Interpolative Decomposition
  Yijun Dong, Chao Chen, Per-Gunnar Martinsson, Katherine Pearce
  Preprint, 2023. [GitHub]
  

• Efficient Bounds and Estimates for Canonical Angles in Randomized Subspace Approximations
  Yijun Dong, Per-Gunnar Martinsson, Yuji Nakatsukasa
  SIAM Journal on Matrix Analysis and Applications, 2024. [GitHub]
  

• Cluster-aware Semi-supervised Learning: Relational Knowledge Distillation Provably Learns Clustering
  Yijun Dong*, Kevin Miller*, Qi Lei, Rachel Ward
  Neural Information Processing Systems (NeurIPS), 2023. [GitHub]
  

• Sample Efficiency of Data Augmentation Consistency Regularization
  Shuo Yang*, Yijun Dong*, Rachel Ward, Inderjit Dhillon, Sujay Sanghavi, Qi Lei
  International Conference on Artificial Intelligence and Statistics (AISTATS), 2023. [pmlr]
  

• Simpler is better: A comparative study of randomized algorithms for computing the CUR decomposition
  Yijun Dong, Per-Gunnar Martinsson
  Advances in Computational Mathematics, 2023. [GitHub]
  

Education

Ph.D. in Computational Science, Engineering, and Mathematics, 2018 - 2023
Oden Institute for Computational Engineering and Sciences, UT Austin, Austin, Texas, US
Thesis: Randomized Dimension Reduction with Statistical Guarantees

B.S. in Applied Mathematics & Engineering Science, 2014 - 2018
Emory University, Atlanta, Georgia, US
Thesis: Crystals and Liquids in Gravitationally Confined Quasi-2-Dimensional Colloidal Systems