Selected papers organized by topic back to homepage

RandNLA Robust solvers Tensors NMF RMT SGD Other

Randomized NLA:

  • Randomized Kaczmarz methods with beyond-Krylov convergence
    ... with M. Derezinski, D. Needell, J. Yang
    ... (2025) submitted, arXiv:2501.11673
  • Fine-grained analysis and faster algorithms for iteratively solving linear systems
    ... with M. Derezinski, D. LeJeune, D. Needell
    ... (2024) submitted, arXiv:2405.05818
  • A Subspace Constrained Randomized Kaczmarz method for structure or external knowledge exploitation
    ... with J. Lok
    ... (2023) Linear Algebra and its Applications (2024), or arXiv:2309.04889
  • Sharp analysis of sketch-and-project methods via a connection to randomized singular value decomposition
    ... with M. Derezinski
    ... (2023) SIAM Journal on Mathematics of Data Science 6 (1), 127-153, or arXiv:2208.09585
  • On block Gaussian sketching for the Kaczmarz method
    ... with D. Needell
    ... (2020) Numerical Algorithms 86 (1), 443-473, or arXiv:1905.08894
  • Sketching for Motzkin's iterative method for linear systems
    ... with D. Needell
    ... (2019) Proc. 50th Asilomar Conference (2019), or arXiv:1912.00771

Robust linear solvers:

  • Stochastic gradient descent for streaming linear and rectified linear systems with adversarial corruptions
    ... with H. Jeong, D. Needell
    ... (2024) SIAM Journal on Mathematics of Data Science (SIMODS), to appear, or arXiv:2403.01204
  • On subsampled Quantile Randomized Kaczmarz
    ... with J. Haddock, A. Ma
    ... (2023) Proc. Allerton Conference (2023), or arXiv:2308.07987
  • On block accelerations of Quantile Randomized Kaczmarz for corrupted systems of linear equations
    ... with L. Cheng, B. Jarman, D. Needell
    ... (2022) Inverse Problems 39 (2), 024002, or arXiv:2206.12554
  • Quantile-based iterative methods for corrupted systems of linear equations
    ... with J. Haddock, D. Needell, W. Swartworth
    ... (2022) SIAM Journal on Matrix Analysis and Applications, 43 (2), 605-637, or arXiv:2009.08089

Tensor data compression and recovery:

  • On trimming tensor-structured measurements and efficient low-rank tensor recovery
    ... with S. Suryanarayanan
    ... (2025) submitted, arXiv:2502.02843
  • Fast and low-Memory compressive sensing algorithms for low Tucker-rank tensor approximation from streamed measurements
    ... with C. Haselby, M. Iwen, D. Needell, W. Swartworth
    ... (2023) submitted, arXiv:2308.13709
  • Modewise operators, the tensor restricted isometry property, and low-rank tensor recovery
    ... with C. Haselby, M. Iwen, D. Needell, M. Perlmutter
    ... (2023) Applied and Computational Harmonic Analysis 66, 161-192, or arXiv:2109.10454
  • Lower memory oblivious (tensor) subspace embeddings with fewer random bits: modewise methods for least squares
    ... with M. Iwen, D. Needell, A. Zare
    ... (2021) SIAM Journal on Matrix Analysis and Applications 42 (1), 376-416 or arXiv:1912.08294

Data analysis via matrix and tensor factorizations:

  • Learning nonnegative matrix factorizations from compressed data
    ... with A. Chaudhry
    ... (2024) submitted, arXiv:2409.04994
  • Sparseness-constrained nonnegative tensor factorization for detecting topics at different time scales
    ... with L. Kassab, A. Kryschenko, H.Lyu, D. Molitor, D. Needell, J. Yuan
    ... (2023) Frontiers in Applied Mathematics and Statistics 10, 1287074, or arXiv:2010.01600
  • On a guided Nonnegative Matrix Factorization
    ... with J. Vendrow, J. Haddock, D. Needell
    ... (2021) Proc. ICASSP IEEE Conference (2021), or arXiv:2010.11365
  • COVID-19 literature topic-based search via hierarchical NMF
    ... with R. Grotheer, K. Ha, L. Huang, Y. Huang, A. Kryshchenko, O. Kryshchenko, P. Li, X. Li, D. Needell
    ... (2020) Proc. NLP-COVID19-EMNLP Workshop (2020), or link
  • Analysis of Legal Documents via Non-negative Matrix Factorization Methods
    ... with R. Budahazy, L. Cheng, Y. Huang, A. Johnson, P. Li, J. Vendrow, Z. Wu, D. Molitor, D. Needell
    ... (2021) SCIURO undergraduate journal, or arXiv:2104.14028

Dynamic of SGD

  • On regularization via early stopping for least squares regression
    ... with R. Sonthalia, J. Lok
    ... (2024) submitted, arXiv:2406.04425
  • Discrete error dynamics of mini-batch gradient descent for least squares regression
    ... with J. Lok, R. Sonthalia
    ... (2025) Proc. ALT Conference, or arXiv:2406.03696

Non-asymptotic random matrix theory:

  • Constructive regularization of the random matrix norm
    E. Rebrova
    ... (2020) Journal of Theoretical Probability, 33 (3), 1768-1790, or arXiv:1809.03926
  • Norms of random matrices: local and global problems
    ... E. Rebrova, R. Vershynin
    ... (2018) Advances in Mathematics, 324, 40–83, or arXiv:1608.06953
  • Coverings of random ellipsoids, and invertibility of matrices with i.i.d. heavy-tailed entries
    ... E. Rebrova, K. Tikhomirov
    ... (2018) Israel J. Math., 227 (2), 507-544, or arXiv:1508.06690
  • Spectral Properties of Heavy-Tailed Random Matrices (PhD Thesis)
    ... E. Rebrova
    ... (2018) University of Michigan, see full text and defense presentation

Graph signal processing

  • On graph uncertainty principle and eigenvector delocalization
    ... with P. Salanevich
    ... (2023) Proc. SAMPTA IEEE Conference, link, or arXiv:2306.15810

Scalable linear algebra for machine learning and scientific computing:

  • Scalable and memory-efficient kernel ridge regression
    ... G. Chavez, E. Rebrova, Y. Liu, P. Ghysels, and X. S. Li
    ... (2020) Proc. 34th IPDSP IEEE Symposium, or link
  • A study of clustering techniques and hierarchical matrix formats for kernel ridge regression
    ... E. Rebrova, G. Chavez, Y. Liu, P. Ghysels, and X. S. Li
    ... (2018) IEEE international parallel and distributed processing symposium workshop, or arXiv:1803.10274
  • A parallel hierarchical blocked adaptive cross approximation algorithm
    ... Y. Liu, W. Sid-Lakhdar, E. Rebrova, P. Ghysels, and X. S. Li
    ... (2020) International Journal of High Perf. Comp. Appl., 34 (4), 394-408, or link