Research
Publications
My work spans stochastic and non-convex optimization, reinforcement learning, control, and communication-efficient distributed learning.
19peer-reviewed papers
13conference papers
6journal articles
2current preprints
Current work
Preprints & under review
Global Solutions to Non-Convex Functional Constrained Problems with Hidden Convexity
Oral Presentation Award, NeurIPS 2025 Workshop on Constrained Optimization for Machine Learning
Published & accepted
Peer-reviewed publications
2026
Global Optimality for Constrained Exploration via Penalty Regularization
Oral Presentation Award, European Workshop on Reinforcement Learning 2026
Can SGD Handle Heavy-Tailed Noise?
Oral Presentation Award, NeurIPS 2025 Workshop on Optimization for Machine Learning
2025
Second-order Optimization under Heavy-Tailed Noise: Hessian Clipping and Sample Complexity Limits
Natural Gradient VI: Guarantees for Non-Conjugate Models
Safe-EF: Error Feedback for Non-smooth Constrained Optimization
From Gradient Clipping to Normalization for Heavy-Tailed SGD
Stochastic Optimization under Hidden Convexity
Learning Zero-Sum Linear Quadratic Games with Improved Sample Complexity and Last-Iterate Convergence
Preliminary version presented at CDC 2023.
EF21 with Bells & Whistles: Six Algorithmic Extensions of Modern Error Feedback
2024
Taming Nonconvex Stochastic Mirror Descent with General Bregman Divergence
2023
Momentum Provably Improves Error Feedback!
Two Sides of One Coin: The Limits of Untuned SGD and the Power of Adaptive Methods
Reinforcement Learning with General Utilities: Simpler Variance Reduction and Large State-Action Space
Stochastic Policy Gradient Methods: Improved Sample Complexity for Fisher-non-degenerate Policies
2022
3PC: Three Point Compressors for Communication-Efficient Distributed Training and a Better Theory for Lazy Aggregation
Spotlight presentation, ICML 2022 main track (top 5%)
Sharp Analysis of Stochastic Optimization under Global Kurdyka-Łojasiewicz Inequality
2021
EF21: A New, Simpler, Theoretically Better, and Practically Faster Error Feedback
Oral presentation, NeurIPS 2021 main track (top 1%)
Optimizing Static Linear Feedback: Gradient Method
2020
Use of Projective Coordinate Descent in the Fekete Problem
* Equal contribution.
Looking for citation details?Visit my Google Scholar profile for the latest citation counts and records.
Google Scholar