Sotabase

Career

· Intern, Amazon2025–2025
· Master's Degree, Mathematics, ETH Zurich2020–2021
· PhD Student, ETH Zurich2020–
· Master's Degree, Mathematics, Technical University of Munich2020–2021

Publications (16)

EF21: A New, Simpler, Theoretically Better, and Practically Faster Error Feedback
Neural Information Processing Systems · 2021
182
cited
Optimizing Static Linear Feedback: Gradient Method
SIAM Journal of Control and Optimization · 2020
93
cited
EF21 with Bells & Whistles: Practical Algorithmic Extensions of Modern Error Feedback
arXiv.org · 2021
54
cited
Stochastic Policy Gradient Methods: Improved Sample Complexity for Fisher-non-degenerate Policies
International Conference on Machine Learning · 2023
54
cited
Momentum Provably Improves Error Feedback!
Neural Information Processing Systems · 2023
40
cited
3PC: Three Point Compressors for Communication-Efficient Distributed Training and a Better Theory for Lazy Aggregation
International Conference on Machine Learning · 2022
34
cited
Sharp Analysis of Stochastic Optimization under Global Kurdyka-Lojasiewicz Inequality
Neural Information Processing Systems · 2022
26
cited
Two Sides of One Coin: the Limits of Untuned SGD and the Power of Adaptive Methods
Neural Information Processing Systems · 2023
25
cited
Reinforcement Learning with General Utilities: Simpler Variance Reduction and Large State-Action Space
International Conference on Machine Learning · 2023
14
cited
Taming Nonconvex Stochastic Mirror Descent with General Bregman Divergence
International Conference on Artificial Intelligence and Statistics · 2024
14
cited
Learning Zero-Sum Linear Quadratic Games with Improved Sample Complexity
IEEE Conference on Decision and Control · 2023
8
cited
Stochastic Optimization under Hidden Convexity
SIAM Journal on Optimization · 2023
8
cited
EF21 with Bells&Whistles: Six Algorithmic Extensions of Modern Error Feedback
2021
3
cited
Use of Projective Coordinate Descent in the Fekete Problem
2020
1
cited
BiCompFL: Bi-Directional Compression for Stochastic Federated Learning
Sotabase