Sayantan Choudhury

Postdoctoral Researcher at MBZUAI

Portrait of Sayantan Choudhury

Welcome to my homepage. I am currently a Postdoctoral Researcher at MBZUAI, hosted by Martin Takac and Mladen Kolar. I completed my Ph.D. in Applied Mathematics and Statistics at Johns Hopkins University, where I was very fortunate to work under the supervision of Nicolas Loizou.

I am broadly interested in the mathematical foundations of machine learning, with a particular focus on optimization methods for modern large-scale learning problems.

Before joining Johns Hopkins, I obtained my B.Stat and M.Stat from the Indian Statistical Institute, Kolkata. During Fall 2023, I was a research intern at MBZUAI, where I worked on adaptive methods for modern deep learning.

Contact: sayantan.choudhury4 [at] gmail [dot] com

News

Last Updated: May 12, 2026.

2026 Our paper Muon with Nesterov Momentum: Heavy-Tailed Noise and (Randomized) Inexact Polar Decomposition is now available on arXiv.
2026 Our paper One-shot Conditional Sampling: MMD meets Nearest Neighbors was accepted at ICML 2026.
2025 I started as a Postdoctoral Researcher at MBZUAI in November 2025.
2025 Our paper Multiplayer Federated Learning: Reaching Equilibrium with Less Communication was accepted at NeurIPS 2025.
2025 Our paper Extragradient Method for (L0, L1)-Lipschitz Root-finding Problems was accepted at NeurIPS 2025.
2025 I completed my Ph.D. in Applied Mathematics and Statistics at Johns Hopkins University and received the Chen Family Dissertation Fellowship for Spring 2025.
2025 Methods for Convex (L0, L1)-Smooth Optimization: Clipping, Acceleration, and Adaptivity was accepted at ICLR 2025.
2024 Remove that Square Root: Efficient Scale-Invariant Version of AdaGrad was accepted at NeurIPS 2024, and I presented a poster at NeurIPS 2024 in Vancouver.
2024 I gave an invited talk at Google Research in Bengaluru and received the Acheson J. Duncan Fund for the Advancement of Research in Statistics travel award.
2023 Single-Call Stochastic Extragradient Methods for Structured Non-monotone Variational Inequalities was accepted at NeurIPS 2023, where I also received a Scholar Award.

Selected Publications

  1. ICLR 2025
    Methods for Convex (L0, L1)-Smooth Optimization: Clipping, Acceleration, and Adaptivity
    Eduard Gorbunov, Nazarii Tupitsa, Sayantan Choudhury, Alen Aliev, Peter Richtarik, Samuel Horvath, Martin Takac
    Accepted at ICLR, 2025.
    Paper
  2. ICML 2026
    One-shot Conditional Sampling: MMD meets Nearest Neighbors
    Anirban Chatterjee, Sayantan Choudhury, Rohan Hore
    Accepted at ICML, 2026.
    Paper
  3. Preprint 2026
    Muon with Nesterov Momentum: Heavy-Tailed Noise and (Randomized) Inexact Polar Decomposition
    Sayantan Choudhury, Xiaoran Cheng, Martin Takac, Sen Na, Mladen Kolar
    Preprint, 2026.
    Paper
  4. NeurIPS 2025
    Multiplayer Federated Learning: Reaching Equilibrium with Less Communication
    TaeHo Yoon, Sayantan Choudhury, Nicolas Loizou
    Accepted at NeurIPS, 2025.
    Paper

Education

  • 2025: Ph.D. in Applied Mathematics and Statistics, Johns Hopkins University
  • 2023: MSE in Applied Mathematics and Statistics, Johns Hopkins University
  • 2020: M.Stat, Indian Statistical Institute, Kolkata
  • 2018: B.Stat, Indian Statistical Institute, Kolkata

Research Interests

  • Large-Scale Optimization
  • Adaptive Methods
  • Variational Inequality Problems
  • Federated Learning
  • Min-Max Optimization
  • Distributed and Decentralized Algorithms