Michał Dereziński
Assistant Professor, EECS – Computer Science and Engineering
Studies statistical learning theory and scalable optimization, with a focus on randomized numerical linear algebra and dimensionality reduction for machine learning.
Michał Dereziński's lab at Michigan works on learning theory & optimization. 26 papers since 2023, 16 at top venues, mostly NeurIPS and COLT. 4 PhD students. Applications close .
At a glance
- papers, 2023–2026
- 7/ year
- at T1 venues
- 62%
- PhD students
- 4
- MS students
- 0
- alumni listed
- —
Papers by yearfrom the lab homepage, 2023–2026
- 8 papers2026 · T1 6 · collecting
- 9 papers2025 · T1 4
- 6 papers2024 · T1 5
- 3 papers2023 · T1 1
T1 venuesOtherVenues: NeurIPS, COLT, SODA, STOC, KDD
Research focuspaper titles that use each topic's words
- randomized6 papers
- linear6 papers
- numerical linear2 papers
- linear algebra2 papers
- Learning theory & optimization1 paper
Paper list
From the lab homepage.
20268 papers
- SODA (T1)Approaching Optimality for Solving Dense Linear Systems with Low-Rank Structure (opens in a new tab)
- NeurIPS (T1)Debiasing Random Oblique Projections for Subsampled OLS and Fast CUR in High Dimensions (opens in a new tab)
- Journal of Machine Learning Research 27(162):1-48Have ASkotch: A Neat Solution for Large-scale Kernel Ridge Regression (opens in a new tab)
- COLT (T1)Last-Iterate Convergence of Randomized Kaczmarz and SGD with Greedy Step Size (opens in a new tab)
- SIAM Journal on Matrix Analysis and ApplicationsNumerical Instabilities in the Kaczmarz Method and Stabilization by Iterative Refinement (opens in a new tab)
- SODA (T1)Optimal Subspace Embeddings: Resolving Nelson-Nguyen Conjecture Up to Sub-Polylogarithmic Factors (opens in a new tab)
- NeurIPS (T1)Perfect Parallelization in Mini-Batch SGD with Classical Momentum Acceleration (opens in a new tab)
- COLT (T1)The matrix-vector complexity of Ax=b (opens in a new tab)
20259 papers
- SODA (T1)Faster Linear Systems and Matrix Norm Approximation via Multi-level Sketched Preconditioning (opens in a new tab)
- COLT (T1)Faster Low-Rank Approximation and Kernel Ridge Regression via the Block-Nyström Method (opens in a new tab)
- Journal of Machine Learning Research 26(144):1-49Fine-grained Analysis and Faster Algorithms for Iteratively Solving Linear Systems (opens in a new tab)
- ICALPOptimal Oblivious Subspace Embeddings with Near-optimal Sparsity (opens in a new tab)
- SIAM Journal on Matrix Analysis and Applications 46(4): 2558-2588Randomized Kaczmarz Methods with Beyond-Krylov Convergence (opens in a new tab)
- Journal of Machine Learning Research 26(306), 1-49Second-order Information Promotes Mini-Batch Robustness in Variance-Reduced Gradients (opens in a new tab)
- TMLR (T1)Stochastic Variance-Reduced Newton: Accelerating Finite-Sum Minimization with Large Batches (opens in a new tab)
- SIAM Journal on Matrix Analysis and Applications 46(2):1247-1279Surrogate-based Autotuning for Randomized Sketching Algorithms in Regression Problems (opens in a new tab)
- NeurIPS (T1)Turbocharging Gaussian Process Inference with Approximate Sketch-and-Project (opens in a new tab)
20246 papers
- NeurIPS (T1)Distributed Least Squares in Small Space via Sketching and Bias Reduction (opens in a new tab)
- STOC (T1)Optimal Embedding Dimension for Sparse Subspace Embeddings (opens in a new tab)
- KDD (T1)Recent and Upcoming Developments in Randomized Numerical Linear Algebra for Machine Learning (opens in a new tab)
- SIAM Journal on Mathematics of Data Science 6:1 127-153Sharp Analysis of Sketch-and-Project Methods via a Connection to Randomized Singular Value Decomposition (opens in a new tab)
- STOC (T1)Solving Dense Linear Systems Faster than via Preconditioning (opens in a new tab)
- NeurIPS (T1)Stochastic Newton Proximal Extragradient Method (opens in a new tab)
20233 papers
- COLT (T1)Algorithmic Gaussianization through Sketching: Converting Data into Sub-gaussian Random Designs (opens in a new tab)
- Mathematical Programming 201:473-520Hessian Averaging in Stochastic Newton Methods Achieves Superlinear Convergence (opens in a new tab)
- LAPACK Working NotesRandomized Numerical Linear Algebra - A Perspective on the Field with an Eye to Software (opens in a new tab)
Where alumni went
Alumni are not listed on the lab homepage.
Apply to MichiganFall 2027
Michigan CSE Computer Science PhD
- Round
- Single round
- Opens
- —
- GRE
- See program page
- English
- See program page
- Fee
- See program page
Not yet confirmed for this intake.
Similar labsnearest by research summary
- CNSJTU
Chenglin Li
- Reinforcement learning
- CHETH Zurich
Niao He
- Learning theory & optimization
- CNTsinghua
Yu Liu
- Reinforcement learning
- GBEdinburgh
Timothy Hospedales
- Efficient training & inference
- GBOxford
Yarin Gal
- Probabilistic & causal inference
- SGNTU
Lihui Chen
- Recognition, detection & segmentation
Your notes
Save this lab to keep notes.
Data
Some numbers could not be verified:
- The alumni list could not be found on the lab site.
- Collected
- Verified
- Papers
- The lab homepage
- Members
- Lab homepage
- Alumni
- Not collected
Tiers are venue classifications, not lab rankings.
Michigancloses