Best Application Paper Award at MOTOR 2026
Our paper Last Iterate Convergence of AdaGrad-Norm for Convex Non-Smooth Optimization received the Best Application Paper Award at MOTOR 2026.
News
2026 conference acceptances
My collaborators and I had several papers accepted at 2026 conferences: one at AISTATS, one at ICLR, five at ICML, two at CPAL, and two at UAI.
- AISTATS 20261 paperDifferentially Private Clipped-SGD
- ICLR 20261 paperHigh-Probability Bounds for the Last Iterate of Clipped SGD
- ICML 20265 papersOptimization to Generalization under Heavy-Tailed DataGeneral Analysis of LMO-based OptimizersAnchored Generalized Optimistic MethodBatch Size in Stochastic Conditional Gradient MethodsBatch Noise, Adaptivity, and Compression under $(L_0,L_1)$-Smoothness
- CPAL 20262 papersByzantine-Robust Optimization under $(L_0,L_1)$-SmoothnessSelective Collaboration for Robust Federated Learning
- UAI 20262 papersByzantine-Robust and Differentially Private Federated OptimizationWho to Trust? Federated Distillation
New preprint on differentially private Clipped-SGD
A new preprint studies high-probability convergence with an arbitrary clipping level.
ICLR and ICML 2025 papers, new preprints, and MBZUAI appointment
- Two of my papers were accepted to ICLR 2025: first-order methods for convex $(L_0,L_1)$-smooth optimization and random reshuffling with local steps under $(L_0,L_1)$-smoothness.
- One paper on high-probability convergence of AdaGrad/Adam-type methods under heavy-tailed noise was accepted to ICML 2025.
- I added several new preprints to the Publications page.
- I started as an Assistant Professor in the Department of Statistics and Data Science at MBZUAI on August 1, 2025.
Four NeurIPS 2024 papers and a new preprint
- Federated learning with random reshuffling, compression, and local steps.
- Byzantine robustness with partial participation.
- A scale-invariant version of AdaGrad.
- Second-order and quasi-Newton methods for variational inequalities (spotlight).
- New preprint: Methods for Convex $(L_0,L_1)$-Smooth Optimization: Clipping, Acceleration, and Adaptivity.
Paper accepted to EMNLP 2024 Findings
The paper Low-Resource Machine Translation through the Lens of Personalized Federated Learning, joint with Viktor Moskvoretskii, Nazarii Tupitsa, Chris Biemann, Samuel Horvath, and Irina Nikishina, was accepted to EMNLP 2024 Findings.
Research Scientist appointment, conference updates, and a new preprint
- I started as a Research Scientist at MBZUAI on April 1, 2024.
- AISTATS 2024 presentations covered heavy-tailed noise and Byzantine robust learning.
- One paper was accepted to ICML 2024 for oral presentation, and a new preprint on Jacobian inexactness in variational inequalities was posted.
New preprint on scale-invariant AdaGrad
Remove that Square Root: A New Efficient Scale-Invariant Version of AdaGrad.
New preprints, a talk, and NeurIPS 2023 papers
Seven new preprints were added, including work on intermediate gradient methods, heavy-tailed stochastic minimization, Byzantine robust learning, and distributed variational inequalities.
- Talk at TES Conference on Mathematical Optimization for Machine Learning.
- Three papers were accepted to NeurIPS 2023 on stochastic extragradient methods, accelerated zeroth-order methods, and Byzantine-tolerant distributed variational inequalities.
Earlier updates8 updates, 2022-2023
New preprints, Hamburg talk, and ICML posters
- Three new preprints covered heavy-tailed bandits, partial personalization in federated learning, and distributed gradient clipping.
- Talk at the LT Group Oberseminar, University of Hamburg, on heavy-tailed stochastic optimization.
- ICML posters were added to the talks and posters archive.
ICML 2023 papers
Two papers were accepted to ICML 2023.
- Convergence beyond monotonicity for proximal point and extragradient-based methods.
- High-probability bounds for stochastic optimization and variational inequalities with unbounded variance.
AISTATS, ICLR, PEP talk, and new preprints
- One paper was accepted to AISTATS 2023 and one paper was accepted to ICLR 2023.
- Talk at PEP talks in Belgium.
- Five new preprints were posted, including work on stochastic variational inequalities and Byzantine-robust optimization.
Preprint on negative comonotonicity
New preprint on convergence of proximal point and extragradient-based methods beyond monotonicity.
Talk at MBZUAI Collaborative Learning Workshop
Presentation on variance reduction for Byzantine-robust distributed optimization.
Outstanding Reviewer Award, NeurIPS 2022
Outstanding Reviewer Award at NeurIPS 2022.
NeurIPS 2022 papers
Two papers were accepted to NeurIPS 2022.
- Clipped stochastic methods for variational inequalities with heavy-tailed noise.
- Last-iterate convergence of optimistic gradient method for monotone variational inequalities.
Postdoctoral position at MBZUAI
I started as a postdoctoral researcher at MBZUAI in Abu Dhabi, in the groups of Samuel Horváth and Martin Takáč.