Anna Ma

My research focuses on problems arising in mathematical data science. I am particularly interested in designing and analyzing iterative algorithms for large-scale data and in developing tools from numerical linear algebra, signal processing, machine learning, and probability.

Below, I highlight some recent works and provide code, when available, associated with the manuscript.

A complete list of my publications can be found on Google Scholar.

Preprints

• Quantile-based Loss Filtering for Outlier-Robust Stochastic Gradient Descent
  J. Haddock, A. Ma, E. Rebrova, 2026.
  arXiv · Code

• Quantile Randomized Kaczmarz for Streaming Linear Systems with Massart Noise
  E. Battaglia, J.-F. Cai, J. Chen, A. Ma, D. Needell, and T. Wu, 2026.
  arXiv · Code

• Attention Mechanisms Through the Lens of Numerical Methods: Approximation Methods and Alternative Formulations
  M. F. Serret, A. Cortinovis, Y. Dong, D. Halikias, A. Ma, F. Matti, D. Needell, K. J. Pearce, E. Rebrova, D. Shur, R. Smith, H.-X. Wang, and L. Grigori, 2026.
  arXiv · Code

• Wedge Sampling: Efficient Tensor Completion with Nearly-Linear Sample Complexity
  H. Luo, A. Ma, L. Stephan, and Y. Zhu, 2026.
  arXiv

• Where Have All the Kaczmarz Iterates Gone?
  E. H. Bergou, S. Boucherouite, A. Dutta, X. Li, and A. Ma, 2025.
  arXiv · Code

• Efficient and Robust Bayesian Selection of Hyperparameters in Dimension Reduction for Visualization
  Y.-T. Liao, H. Luo, and A. Ma, 2023.
  arXiv

Research Areas

Randomized Iterative Methods

Randomized iterative methods solve large-scale linear systems through a sequence of inexpensive updates based on randomly selected equations, blocks, or sketches. This approach can reduce memory and computational costs while remaining effective when data are noisy, inconsistent, or corrupted.

Paper highlights

• On the Subsample Size of Quantile-Based Randomized Kaczmarz
  J.-F. Cai, J. Chen, A. Ma, and T. Wu.
  arXiv · SIAM Journal on Matrix Analysis and Applications · Code

• Quantile-RK and Double Quantile-RK Error Horizon Analysis
  E. Battaglia and A. Ma.
  arXiv · Linear Algebra and its Applications · Code

• Greed Works: An Improved Analysis of Sampling Kaczmarz-Motzkin
  J. Haddock and A. Ma.
  arXiv · SIAM Journal on Mathematics of Data Science

Tensor Methods

Tensor methods preserve the multiway structure found in data such as images, videos, and scientific measurements. Low-rank models, structured sampling, and tensor linear algebra make it possible to recover and process these datasets efficiently, even when observations are incomplete or noisy.

Paper highlights

• Wedge Sampling: Efficient Tensor Completion with Nearly-Linear Sample Complexity
  H. Luo, A. Ma, L. Stephan, and Y. Zhu.
  arXiv · Conference on Learning Theory (COLT)

• Stochastic Gradient Descent for Incomplete Tensor Linear Systems
  A. Ma, D. Needell, and A. Xue.
  arXiv · BIT Numerical Mathematics · Code

• Robust Recovery of Low-Rank Matrices and Low-Tubal-Rank Tensors from Noisy Sketches
  A. Ma, D. Stöger, and Y. Zhu.
  arXiv · SIAM Journal on Matrix Analysis and Applications · Code

Data Visualization and Machine Learning

Data visualization transforms high-dimensional information into interpretable low-dimensional representations, while machine learning identifies structure and patterns within complex datasets. Numerical linear algebra and optimization provide the tools needed to make these methods scalable, stable, and efficient.

Paper highlights

• Attention Mechanisms Through the Lens of Numerical Methods: Approximation Methods and Alternative Formulations
  M. F. Serret, A. Cortinovis, Y. Dong, D. Halikias, A. Ma, F. Matti, D. Needell, K. J. Pearce, E. Rebrova, D. Shur, R. Smith, H.-X. Wang, and L. Grigori.
  arXiv · Code

• Efficient and Robust Bayesian Selection of Hyperparameters in Dimension Reduction for Visualization
  Y.-T. Liao, H. Luo, and A. Ma.
  arXiv

• AVIDA: An Alternating Method for Visualizing and Integrating Data
  K. Dover, Z. Cang, A. Ma, Q. Nie, and R. Vershynin.
  arXiv · Journal of Computational Science · Code

Last updated September 2026

Powered by Jekyll and Minimal Light theme.