Reading

• The course will closely follow lecture notes prepared by the instructor:

  • Lecture Notes for ESE6180 by Ingvar Ziemann

    • Please note that these are a work in progress and will be updated throughout the Course
    • If you find any typos or other mistakes, an email to ingvarz@seas.upenn.edu will be much appreciated.

• The following may also be helpful

  • Our 2022 Survey of the Statistical Learning for Control

  • High-Dimensional Statistics by Martin Wainwright

  • High-Dimensional Probability by Roman Vershynin. NB: freely available at <—

  • Our 2023 Tutorial on the Non-Asymptotic Theory of System Identification

  • High-Dimensional Statistics Lecture notes by Philippe Rigollet and Philippe Rigollet and Jan-Christian Hütter

  • Link to previous year’s course

Suggested Further Reading

---

The following papers are suitable for further reading (and the lectures are loosely based on a number of them).

• Probability, Machine Learning Theory and Statistics:
  • Hanson-Wright inequality and sub-gaussian concentration, Mark Rudelson and Roman Vershynin
  • Random Design Analysis of Ridge Regression, Daniel Hsu, Sham Kakade and Tong Zhang
  • Online Least Squares Estimation with Self-Normalized Processes, Yasin Abassi-Yadkori, David Pal, Csaba Szepesvari
• System Identification:
  • Learning without Mixing, Max Simchowitz, Horia Mania, Stephen Tu, Michael Jordan and Benjamin Recht
  • Sample Complexity Lower Bounds for Linear System Identification, Yassir Jedra and Alexandre Proutiere
  • Revisiting Ho–Kalman-Based System Identification: Robustness and Finite-Sample Analysis, Samet Oymak and Necmiye Ozay
  • Finite Sample Analysis of Stochastic System Identification, Anastasios Tsiamis and George J. Pappas
  • Finite Sample Frequency Domain Identification, Anastasios Tsiamis, Mohamed Abdoalmaty, Roy S. Smith and John Lygeros
• Learning in LQR:
  • Certainty equivalence is efficient for linear quadratic control, Horia Mania, Stephen Tu and Benjamin Recht
  • Global convergence of policy gradient methods for the linear quadratic regulator, Maryam Fazel, Rong Ge, Sham Kakade and Mehran Mesbahi
  • Naive exploration is optimal for online LQR, Max Simchowitz and Dylan Foster
  • Task-Optimal Exploration in Linear Dynamical Systems, Andrew Wagenmaker, Max Simchowitz and Kevin Jamieson
  • Regret Lower Bounds for Linear Quadratic Gaussian Systems, Ingvar Ziemann and Henrik Sandberg
  • Improper Learning for Non-Stochastic Control, Max Simchowitz, Karan Singh and Elad Hazan
• Learning in Nonlinear Systems:
  • Learning with little mixing, Ingvar Ziemann and Stephen Tu
  • Near-optimal offline and streaming algorithms for learning non-linear dynamical systems, Prateek Jain, Suhas Kowshik, Dheeraj Nagaraj and Praneeth Netrapalli
  • Sharp Rates in Dependent Learning Theory, Ingvar Ziemann, Stephen Tu, George J. Pappas and Nikolai Matni
• Learning for Nonlinear Control
  • Information Theoretic Regret Bounds for Online Nonlinear Control, Sham Kakade, Akshay Krishnamurthy, Kendall Lowrey, Motoya Ohnishi, Wen Sun
  • Toward the Fundamental Limits of Imitation Learning, Nived Rajaraman, Lin F. Yang, Jiantao Jiao, Kannan Ramachandran
  • TaSIL: Taylor Series Imitation Learning, Daniel Pfrommer, Thomas Zhang, Stephen Tu, Nikolai Matni
  • Active Learning for Control-Oriented Identification of Nonlinear Systems, Bruce Lee, Ingvar Ziemann, George Pappas and Nikolai Matni
  • Is Behavior Cloning All You Need? Understanding Horizon in Imitation Learning, Dylan J. Foster, Adam Block, Dipendra Misra