Stephen Tu

I am an Assistant Professor in the Department of Electrical and Computer Engineering at the University of Southern California.

I am currently recruiting PhD students to work with me on problems in controls, robotics, generative modeling, and learning from dynamical systems!

If my research interests align well with yours, please do not hesitate to reach out.


The easiest way to reach me is e-mail: stephen.tu@usc.edu

jax4dc tutorial

Roy Frostig, Sumeet Singh, and myself recently gave a jax4dc tutorial at L4DC 2023. This was a fun excursion into leveraging the power of functional programming and automatic differentiation for dynamics and control problems. Please check out the slides and Jupyter notebooks!

Teaching

EE660, Machine Learning II: Mathematical Foundations and Methods (Spring 2024, Fall 2024).

Preprints

Shallow diffusion networks provably learn hidden low-dimensional structure. [Paper]
Nicholas M. Boffi, Arthur Jacot, Stephen Tu, and Ingvar Ziemann.

Incremental Composition of Learned Control Barrier Functions in Unknown Environments. [Paper]
Paul Lutkus, Deepika Anantharaman, Stephen Tu, and Lars Lindemann.

Publications

Revisiting Energy Based Models as Policies: Ranking Noise Contrastive Estimation and Interpolating Energy Models. [Paper]
Sumeet Singh, Stephen Tu, and Vikas Sindhwani.
Transactions on Machine Learning Research, 2024.

Learning from many trajectories. [Paper]
Stephen Tu, Roy Frostig, and Mahdi Soltanolkotabi.
Journal of Machine Learning Research, Vol. 25, No. 216, 2024.

Learning Robust Output Control Barrier Functions from Safe Expert Demonstrations. [Paper]
Lars Lindemann, Alexander Robey, Lejun Jiang, Satyajeet Das, Stephen Tu, and Nikolai Matni.
IEEE Open Journal of Control Systems, Vol. 3, 2024.

Sharp Rates in Dependent Learning Theory: Avoiding Sample Size Deflation for the Square Loss. [Paper]
Ingvar Ziemann, Stephen Tu, George J. Pappas, and Nikolai Matni.
ICML 2024.

The noise level in linear regression with dependent data. [Paper]
Ingvar Ziemann, Stephen Tu, George J. Pappas, and Nikolai Matni.
NeurIPS 2023.

Robots That Ask For Help: Uncertainty Alignment for Large Language Model Planners. [Paper]
Allen Z. Ren, Anushri Dixit, Alexandra Bodrova, Sumeet Singh, Stephen Tu, Noah Brown, Peng Xu, Leila Takayama, Fei Xia, Jake Varley, Zhenjia Xu, Dorsa Sadigh, Andy Zeng, and Anirudha Majumdar.
CoRL 2023, Best Student Paper.

Bootstrapped Representations in Reinforcement Learning. [Paper]
Charline Le Lan, Stephen Tu, Mark Rowland, Anna Harutyunyan, Rishabh Agarwal, Marc G. Bellemare, and Will Dabney.
ICML 2023.

The Power of Learned Locally Linear Models for Nonlinear Policy Optimization. [Paper]
Daniel Pfrommer, Max Simchowitz, Tyler Westenbroek, Nikolai Matni, and Stephen Tu.
ICML 2023.

Agile Catching with Whole-Body MPC and Blackbox Policy Learning. [Paper]
Saminda Abeyruwan, Alex Bewley, Nicholas M. Boffi, Krzysztof Choromanski, David D'Ambrosio, Deepali Jain, Pannag Sanketi, Anish Shankar, Vikas Sindhwani, Sumeet Singh, Jean-Jacques Slotine, and Stephen Tu.
L4DC 2023.

Multi-Task Imitation Learning for Linear Dynamical Systems. [Paper]
Thomas T. Zhang, Katie Kang, Bruce D. Lee, Claire Tomlin, Sergey Levine, Stephen Tu, and Nikolai Matni.
L4DC 2023.

Visual Backtracking Teleoperation: A Data Collection Protocol for Offline Image-Based Reinforcement Learning. [Paper]
David Brandfonbrener, Stephen Tu, Avi Singh, Stefan Welker, Chad Boodoo, Nikolai Matni, and Jake Varley.
ICRA 2023.

Learning Model Predictive Controllers with Real-Time Attention for Real-World Navigation. [Paper]
Xuesu Xiao, Tingnan Zhang, Krzysztof Choromanski, Edward Lee, Anthony Francis, Jake Varley, Stephen Tu, Sumeet Singh, Peng Xu, Fei Xia, Sven Mikael Persson, Dmitry Kalashnikov, Leila Takayama, Roy Frostig, Jie Tan, Carolina Parada, and Vikas Sindhwani.
CoRL 2022.

Learning with little mixing. [Paper]
Ingvar Ziemann and Stephen Tu.
NeurIPS 2022.

TaSIL: Taylor Series Imitation Learning. [Paper]
Daniel Pfrommer, Thomas T.C.K. Zhang, Stephen Tu, and Nikolai Matni.
NeurIPS 2022.

Adversarially Robust Stability Certificates can be Sample-Efficient. [Paper]
Thomas T.C.K. Zhang, Stephen Tu, Nicholas M. Boffi, Jean-Jacques E. Slotine, and Nikolai Matni.
L4DC 2022.

On the Sample Complexity of Stability Constrained Imitation Learning. [Paper]
Stephen Tu, Alexander Robey, Tingnan Zhang, and Nikolai Matni.
L4DC 2022.

The role of optimization geometry in single neuron learning. [Paper]
Nicholas M. Boffi, Stephen Tu, and Jean-Jacques E. Slotine.
AISTATS 2022.

On the Generalization of Representations in Reinforcement Learning. [Paper]
Charline Le Lan, Stephen Tu, Adam Oberman, Rishabh Agarwal, and Marc G. Bellemare.
AISTATS 2022.

Nonparametric adaptive control and prediction: theory and randomized algorithms. [Paper]
Nicholas M. Boffi, Stephen Tu, and Jean-Jacques E. Slotine.
Journal of Machine Learning Research, Vol. 23, No. 281, 2022.
(Earlier version appeared at CDC 2021).

Learning Robust Hybrid Control Barrier Functions for Uncertain Systems. [Paper]
Alexander Robey, Lars Lindemann, Stephen Tu, and Nikolai Matni.

ADHS 2021.

Regret Bounds for Adaptive Nonlinear Control. [Paper]
Nicholas M. Boffi*, Stephen Tu*, and Jean-Jacques E. Slotine.
* Equal contribution.
L4DC 2021.

Safely Learning Dynamical Systems from Short Trajectories. [Paper]
Amir Ali Ahmadi, Abraar Chaudhry, Vikas Sindhwani, and Stephen Tu.
L4DC 2021.

Learning Hybrid Control Barrier Functions from Data. [Paper]
Lars Lindemann, Haimin Hu, Alexander Robey, Hanwen Zhang, Dimos V. Dimarogonas, Stephen Tu, and Nikolai Matni.
CoRL 2020.

Learning Stability Certificates from Data. [Paper]
Nicholas M. Boffi*, Stephen Tu*, Nikolai Matni, Jean-Jacques E. Slotine, and Vikas Sindhwani.
* Equal contribution.
CoRL 2020.

Learning Control Barrier Functions from Expert Demonstrations. [Paper]
Alexander Robey, Haimin Hu, Lars Lindemann, Hanwen Zhang, Dimos V. Dimarogonas, Stephen Tu, and Nikolai Matni.
CDC 2020.

Observational Overfitting in Reinforcement Learning. [Paper]
Xingyou Song, Yiding Jiang, Stephen Tu, Yilun Du, and Behnam Neyshabur.
ICLR 2020.

From Self-Tuning Regulators to Reinforcement Learning and Back Again. [Paper]
Nikolai Matni, Alexandre Proutiere, Anders Rantzer, and Stephen Tu.
CDC 2019.

A Tutorial on Concentration Bounds for System Identification. [Paper]
Nikolai Matni and Stephen Tu.
CDC 2019.

Finite-time Analysis of Approximate Policy Iteration for the Linear Quadratic Regulator. [Paper]
Karl Krauth*, Stephen Tu*, and Benjamin Recht.
* Equal contribution.
NeurIPS 2019.

Certainty Equivalence is Efficient for Linear Quadratic Control. [Paper]
Horia Mania, Stephen Tu, and Benjamin Recht.
NeurIPS 2019.

The Gap Between Model-Based and Model-Free Methods on the Linear Quadratic Regulator: An Asymptotic Viewpoint. [Paper]
Stephen Tu and Benjamin Recht.
COLT 2019.

On the Sample Complexity of the Linear Quadratic Regulator. [Paper]
Sarah Dean, Horia Mania, Nikolai Matni, Benjamin Recht, and Stephen Tu.
Foundations of Computational Mathematics, Vol. 20, 2020.

Minimax Lower Bounds for H-Infinity-Norm Estimation. [Paper]
Stephen Tu*, Ross Boczar*, and Benjamin Recht.
* Equal contribution.
ACC 2019.

Safely Learning to Control the Constrained Linear Quadratic Regulator. [Paper]
Sarah Dean, Stephen Tu, Nikolai Matni, and Benjamin Recht.
ACC 2019.

Regret Bounds for Robust Adaptive Control of the Linear Quadratic Regulator. [Paper]
Sarah Dean, Horia Mania, Nikolai Matni, Benjamin Recht, and Stephen Tu.
NeurIPS 2018.

Least-Squares Temporal Difference Learning for the Linear Quadratic Regulator. [Paper]
Stephen Tu and Benjamin Recht.
ICML 2018.

Learning Without Mixing: Towards A Sharp Analysis of Linear System Identification. [Paper]
Max Simchowitz, Horia Mania, Stephen Tu, Michael I. Jordan, and Benjamin Recht.
COLT 2018.

On the Approximation of Toeplitz Operators for Nonparametric H-infinity-norm Estimation. [Paper]
Stephen Tu, Ross Boczar, and Benjamin Recht.
ACC 2018.

Breaking Locality Accelerates Block Gauss-Seidel. [Paper] [Slides]
Stephen Tu, Shivaram Venkataraman, Ashia C. Wilson, Alex Gittens,
Michael I. Jordan, and Benjamin Recht.
ICML 2017.

Cyclades: Conflict-free Asynchronous Machine Learning. [Paper]
Xinghao Pan, Maximilian Lam, Stephen Tu, Dimitris Papailiopoulos, Ce Zhang, Michael I. Jordan, Kannan Ramchandran, Christopher Ré, and Benjamin Recht.
NeurIPS 2016.

Low-rank Solutions of Linear Matrix Equations via Procrustes Flow. [Paper] [Slides]
Stephen Tu, Ross Boczar, Max Simchowitz, Mahdi Soltanolkotabi, and Benjamin Recht.
ICML 2016.

Machine Learning Classification over Encrypted Data. [Paper]
Raphael Bost, Raluca Ada Popa, Stephen Tu, and Shafi Goldwasser.
NDSS 2015.

Fast Databases with Fast Durability and Recovery through Multicore Parallelism. [Paper]
Wenting Zheng, Stephen Tu, Eddie Kohler, and Barbara Liskov.
OSDI 2014.

Anti-Caching: A New Approach to Swapping in Main Memory OLTP Database Systems. [Paper]
Justin DeBrabant, Andrew Pavlo, Stephen Tu, Michael Stonebraker, and Stan Zdonik.
VLDB 2014.

Speedy Transactions in Multicore In-Memory Databases. [Paper] [Slides] [Code]
Stephen Tu, Wenting Zheng, Eddie Kohler, Barbara Liskov, and Samuel Madden.
SOSP 2013.

Processing Analytical Queries over Encrypted Data. [Paper] [Slides] [Code]
Stephen Tu, M. Frans Kaashoek, Samuel Madden, and Nickolai Zeldovich.
VLDB 2013.

The HipHop Compiler for PHP.
Haiping Zhao, Iain Proctor, Minghui Yang, Xin Qi, Mark Williams, Guilherme Ottoni, Charlie Gao, Andrew Paroski, Scott MacVicar, Jason Evans, and Stephen Tu.
OOPSLA 2012.

The Case for PIQL: A Performance Insightful Query Language.
Michael Armbrust, Nick Lanham, Stephen Tu, Armando Fox, Michael Franklin, and David Patterson.
SoCC 2010.

PIQL: A Performance Insightful Query Language For Interactive Applications.
Michael Armbrust, Stephen Tu, Armando Fox, Michael Franklin, David Patterson, Nick Lanham, Beth Trushkowsky, and Jesse Trutna.
SIGMOD 2010, Demonstration.

Old preprints

Learning Contracting Vector Fields For Stable Imitation Learning. [arXiv]
Vikas Sindhwani, Stephen Tu, and Mohi Khansari.

Non-Asymptotic Analysis of Robust Control from Coarse-Grained Identification. [arXiv]
Stephen Tu, Ross Boczar, Andrew Packard, and Benjamin Recht.

Large Scale Kernel Learning using Block Coordinate Descent. [arXiv]
Stephen Tu, Rebecca Roelofs, Shivaram Venkataraman, and Benjamin Recht.

Thesis

Sample Complexity Bounds for the Linear Quadratic Regulator. [PDF]
PhD Thesis, University of California, Berkeley. Spring 2019.

Writings

An elementary proof of anti-concentration for degree two non-negative Gaussian polynomials (with Ross Boczar). [PDF]

On the exponential convergence of Langevin diffusions. [PDF]

Learning and Control with Safety and Stability Guarantees for Nonlinear Systems. [PDF]

On the Smallest Singular Value of Non-Centered Gaussian Designs. [PDF]

Learning mixture models. [PDF]

Practical first order methods for large scale semidefinite programming. [PDF]

Geometric random walks for sampling from convex bodies. [PDF]

data-microscopes: Bayesian non-parametric inference made simple in Python. [Slides]

The Dirichlet-Multinomial and Dirichlet-Categorical models for Bayesian inference. [PDF]

Derivation of EM updates for discrete Hidden Markov Models. [PDF]

Techniques for query processing on encrypted databases. [PDF]

Implementing concurrent data structures on modern multicore machines. [Slides] [Examples]