Research Group Suvrit Sra
Suvrit Sra
is Professor for Resource Aware Machine Learning at TU Munich.
He specializes in robust, reliable, and resource-efficient machine learning methods. His research focuses, in particular, on solving optimization problems for machine learning with multiple parameters. For example, these complex optimization problems are used in autonomous driving so that a car can reliably distinguish a sign from a person.
Team members @MCML
PostDocs
PhD Students
Recent News @MCML
Publications @MCML
2026
[11]
R. Srivastava • P. Fatemi • S. Sra
BoundKD: Distillation as Constraint Satisfaction under Sampled Supervision.
LIGHT @NeurIPS 2026 - Workshop on Deployable Small Foundation Models at the 40th Conference on Neural Information Processing Systems. Sydney, Australia, Dec 06-12, 2026. To be published. Preprint available. arXiv
BoundKD: Distillation as Constraint Satisfaction under Sampled Supervision.
LIGHT @NeurIPS 2026 - Workshop on Deployable Small Foundation Models at the 40th Conference on Neural Information Processing Systems. Sydney, Australia, Dec 06-12, 2026. To be published. Preprint available. arXiv
[10]
P. Fatemi • H. Maskan • S. Sra • P. Esfahani
Tight Generalization Bounds for Noiseless Inverse Optimization.
NeurIPS 2026 - 40th Conference on Neural Information Processing Systems. Sydney, Australia, Dec 06-12, 2026. To be published. Preprint available. arXiv
Tight Generalization Bounds for Noiseless Inverse Optimization.
NeurIPS 2026 - 40th Conference on Neural Information Processing Systems. Sydney, Australia, Dec 06-12, 2026. To be published. Preprint available. arXiv
[9]
P. Fatemi • H. Maskan • A. Yurtsever • S. Sra
The Multi-Block DC Function Class: Theory, Algorithms, and Applications.
NeurIPS 2026 - 40th Conference on Neural Information Processing Systems. Sydney, Australia, Dec 06-12, 2026. To be published. Preprint available. arXiv
The Multi-Block DC Function Class: Theory, Algorithms, and Applications.
NeurIPS 2026 - 40th Conference on Neural Information Processing Systems. Sydney, Australia, Dec 06-12, 2026. To be published. Preprint available. arXiv
[8]
Z. Liu • P. Fatemi • S. Sra
Transporting the Past: An Optimal Transport View of Backtracking Counterfactuals.
NeurIPS 2026 - 40th Conference on Neural Information Processing Systems. Sydney, Australia, Dec 06-12, 2026. To be published. Preprint available. arXiv
Transporting the Past: An Optimal Transport View of Backtracking Counterfactuals.
NeurIPS 2026 - 40th Conference on Neural Information Processing Systems. Sydney, Australia, Dec 06-12, 2026. To be published. Preprint available. arXiv
[7]
S. N. Ramachandran • S. Sra
Percolation Dynamics in Optimization : Variance Cascades and Discrete Scale Invariance.
Preprint (Sep. 2026). arXiv
Percolation Dynamics in Optimization : Variance Cascades and Discrete Scale Invariance.
Preprint (Sep. 2026). arXiv
[6]
S. N. Ramachandran • S. Sra
Trees to Flows and Back: Unifying Decision Trees and Diffusion Models.
ICML 2026 - 43rd International Conference on Machine Learning. Seoul, South Korea, Jul 06-11, 2026. To be published. Preprint available. URL
Trees to Flows and Back: Unifying Decision Trees and Diffusion Models.
ICML 2026 - 43rd International Conference on Machine Learning. Seoul, South Korea, Jul 06-11, 2026. To be published. Preprint available. URL
[5]
F. Hübler • T. Pethick • S. Sra
Free Heavy-Tailed Lunch for Muon: A Theoretical Justification of Empirical Success.
Preprint (Jun. 2026). arXiv GitHub
Free Heavy-Tailed Lunch for Muon: A Theoretical Justification of Empirical Success.
Preprint (Jun. 2026). arXiv GitHub
[4]
M.-L. Vladarean • X. Zhang • S. Sra
On learning linear dynamical systems in context with attention layers.
ICLR 2026 - 14th International Conference on Learning Representations. Rio de Janeiro, Brazil, Apr 23-27, 2026. To be published. Preprint available. URL
On learning linear dynamical systems in context with attention layers.
ICLR 2026 - 14th International Conference on Learning Representations. Rio de Janeiro, Brazil, Apr 23-27, 2026. To be published. Preprint available. URL
2025
[3]
S. N. Ramachandran • M. K. Lal • S. Sra
Cross-fluctuation phase transitions reveal sampling dynamics in diffusion models.
NeurIPS 2025 - 39th Conference on Neural Information Processing Systems. San Diego, CA, USA, Nov 30-Dec 07, 2025. URL
Cross-fluctuation phase transitions reveal sampling dynamics in diffusion models.
NeurIPS 2025 - 39th Conference on Neural Information Processing Systems. San Diego, CA, USA, Nov 30-Dec 07, 2025. URL
[2]
P. Fatemi • E. Sharifian • M. H. Yassaee
A New Approach to Backtracking Counterfactual Explanations: A Unified Causal Framework for Efficient Model Interpretability.
ICML 2025 - 42nd International Conference on Machine Learning. Vancouver, Canada, Jul 13-19, 2025. URL
A New Approach to Backtracking Counterfactual Explanations: A Unified Causal Framework for Efficient Model Interpretability.
ICML 2025 - 42nd International Conference on Machine Learning. Vancouver, Canada, Jul 13-19, 2025. URL
[1]
A. Bergmeister • M. K. Lal • S. Jegelka • S. Sra
A projection-based framework for gradient-free and parallel learning.
Preprint (Jun. 2025). arXiv
A projection-based framework for gradient-free and parallel learning.
Preprint (Jun. 2025). arXiv
©all images: LMU | TUM
Back to Top
2024-12-27 - Last modified: 2026-07-03