Eric Mazumdar

Assistant Professor in Computing and Mathematical Sciences & Economics

I am an Assistant Professor in Computing and Mathematical Sciences and Economics at Caltech. I obtained my Ph.D in Electrical Engineering and Computer Science at UC Berkeley, co-advised by Michael Jordan and Shankar Sastry.

My research interests lie at the intersection of machine learning and economics. I am broadly interested in developing the tools and understanding necessary to confidently deploy machine learning algorithms into societal systems. This requires understanding the theoretical underpinnings of learning algorithms in uncertain, dynamic environments where they interact with strategic agents--- including people and other algorithms. Practically, I apply my work to problems in intelligent infrastructure, online markets, e-commerce, and the delivery of healthcare.

I am the recipient of a NSF Career Award aimed at studying the strategic interactions that arise in Societal-Scale Systems as well as a Research Fellowship for Learning in Games from the Simons Institute for Theoretical Computer Science. My work is supported by NSF, DARPA, and Amazon research grants.

Prior to Berkeley, I received an SB in Electrical Engineering and Computer Science at Massachusetts Institute of Technology (MIT) , where I had the opportunity to work with in the Laboratory for Multiscale Regenerative Technologies as well as in the MIT Computational Biology Group in CSAIL.

Contact

You can contact me by email at: olastnameoatocaltechodotoedu

Office Hours

I keep office hours during academic quarters in my office (ANB 216). I am available during this time for discussions with any students/postdocs who would like to meet. Emailing ahead of time to is preferred in case I am traveling.

Publications

Preprints/Under review

Coupled Wasserstein Gradient Flows for Min-Max and Cooperative Games
Lauren Conger, Franca Hoffman, Eric Mazumdar, Lillian Ratliff

Tractable Equilibrium Computation in Markov Games through Risk Aversion
Eric Mazumdar , Kishan Panaganti, Laixi Shi [PDF]

Flow-Based Synthesis of Reactive Tests for Discrete Decision-Making Systems with Temporal Logic Specifications
Josefine B. Graebener, Apurva S. Badithela, Denizalp Goktas, Wyatt Ubellacker, Eric Mazumdar, Aaron D. Ames, Richard M. Murray
IEEE Transactions of Robotics (Under Submission), 2024 [PDF]

Last-Iterate Convergence of Payoff-Based Independent Learning in Zero-Sum Stochastic Games Zaiwei Chen, Kaiqing Zhang, Eric Mazumdar, Asuman Ozdaglar, Adam Wierman. [PDF]

Refereed Publications

Understanding Model Selection for Learning in Strategic Environments
Tinashe Handina, Eric Mazumdar
Conference on Neural Information Processing Systems (NeurIPS), 2024 [PDF]

Last-Iterate Convergence for Generalized Frank-Wolfe in Monotone Variational Inequalities
Zaiwei Chen, Eric Mazumdar
Conference on Neural Information Processing Systems (NeurIPS), 2024

Near-Optimal Distributionally Robust RL with General Lp Norms
Pierre Clavier, Laixi Shi, Erwan Le Pennec, Eric Mazumdar, Adam Wierman, Matthieu Geist
Conference on Neural Information Processing Systems (NeurIPS), 2024

Characterizing Controllability and Observability for Systems with Locality, Communication, and Actuation Constraints
Lauren Conger, Yiheng Lin, Adam Wierman, Eric Mazumdar
IEEE Conference on Decision and Control (CDC), 2024 [PDF]

Follower Agnostic Learning in Stackelberg Games
Chinmay Maheshwari, James Cheng, Shankar Sastry, Lillian J. Ratliff, Eric Mazumdar
IEEE Conference on Decision and Control (CDC), 2024 [PDF]

Two-Timescale Q-Learning with Function Approximation in Zero-Sum Stochastic Games
Zaiwei Chen, Kaiqing Zhang, Eric Mazumdar, Asuman Özdaglar, Adam Wierman
Conference on Economics and Computation (EC), 2024 [PDF]

Rethinking Scaling Laws for Learning in Strategic Environments
Tinashe Handina, Eric Mazumdar
Econometric Society Interdisciplinary Frontiers (ESIF) conference on Economics and AI+ML, 2024 [PDF]

Model-Free Robust ϕ-Divergence Reinforcement Learning Using Both Offline and Online Data
Kishan Panaganti, Adam Wierman, Eric Mazumdar
International Conference on Machine Learning (ICML), 2024 [PDF]

Sample-Efficient Robust Multi-Agent Reinforcement Learning in the Face of Environmental Uncertainty
Laixi Shi, Eric Mazumdar, Yuejie Chi, Adam Wierman
International Conference on Machine Learning (ICML), 2024 [PDF]

A Finite-Sample Analysis of Payoff-Based Independent Learning in Zero-Sum Stochastic Games
Zaiwei Chen, Kaiqing Zhang, Eric Mazumdar, Asuman Özdaglar, Adam Wierman
Conference on Neural Information Processing Systems (NeurIPS), 2023 [PDF]

Distribution Shifts of Strategic Interacting Agents via Coupled Gradient Flows
Lauren Conger, Franca Hoffman, Eric Mazumdar, Lillian Ratliff
Conference on Neural Information Processing Systems (NeurIPS), 2023 [PDF]

Algorithmic Collective Action in Machine Learning
Moritz Hardt, Eric Mazumdar, Celestine Mendler-Dünner, Tijana Zrnic (α-β ordering) [PDF]
International Conference on Machine Learning (ICML), 2023 [PDF]

Designing System Level Synthesis Controllers for Nonlinear Systems with Stability Guarantees
Lauren Conger, Sydney Vernon, Eric Mazumdar
Conference on Learning for Dynamics and Control (L4DC), 2023

Synthesizing Reactive Test Environments for Autonomous Systems: Testing Reach-Avoid Specifications with Multi-Commodity Flows
Apurva Badithela, Josefine B. Graebener, Wyatt Ubellacker, Eric V. Mazumdar, Aaron D. Ames, Richard M. Murray
International Conference on Robotics and Automation (ICRA), 2023 [PDF]

Decentralized, Coordination- and Communication-Free Algorithms for Learning in Structured Matching Markets
Chinmay Maheshwari, S. Shankar Sastry, Eric Mazumdar
Conference on Neural Information Processing Systems (NeurIPS), 2022 [PDF]

Nonlinear System Level Synthesis for Polynomial Dynamical Systems
Lauren Conger, Jing Shuang (Lisa) Li, Eric Mazumdar, Steven Brunton
IEEE Conference on Decision and Control (CDC), 2022 [PDF]

Langevin Monte Carlo for Contextual Bandits
Pan Xu, Hongkai Zheng, Eric Mazumdar, Kamyar Azizzadenesheli, Anima Anandkumar
International Conference on Machine Learning (ICML), 2022 [PDF]

Fast Distributionally Robust Learning via Min-Max Optimization
Yaodong Yu*, Tianyi Lin*, Eric Mazumdar*, Michael I. Jordan
International Conference on Artificial Intelligence and Statistics (AISTATS), 2022 [PDF]
(* denotes equal contribution)

Zeroth-Order Methods for Convex-Concave Minmax Problems: Applications to Decision-Dependent Risk Minimization
Chinmay Maheshwari, Chih-Yuan Chiu, Eric Mazumdar, S. Shankar Sastry, Lillian J. Ratliff
International Conference on Artificial Intelligence and Statistics (AISTATS), 2022 [PDF]

Who Leads and Who Follows in Strategic Classification?
Tijana Zrnic*, Eric Mazumdar*, S. Shankar Sastry, Michael I. Jordan
Conference on Neural Information Processing Systems (NeurIPS), 2021 [PDF]
(* denotes equal contribution)

Global Convergence to Local Minmax Equilibrium in Classes of Nonconvex Zero-Sum Games
Tanner Fiez, Lillian J. Ratliff, Eric Mazumdar, Evan Faulkner, Adhyyan Narang
Conference on Neural Information Processing Systems (NeurIPS), 2021 [PDF]

High Confidence Sets for Trajectories of Stochastic Time-Varying Nonlinear Systems
Eric Mazumdar, Tyler Westenbroek, Michael I. Jordan, S. Shankar Sastry
IEEE Conference on Decision and Control (CDC), 2020 [PDF]

Adaptive Control for Linearizable Systems Using On-Policy Reinforcement Learning
Tyler Westenbroek, Eric Mazumdar, David Fridovich-Keil, Valmik Prabhu, Claire J. Tomlin, S. Shankar Sastry
IEEE Conference on Decision and Control (CDC), 2020 [PDF]

Expert Selection in High Dimensional Markov Decision Processes
Vicenc Rubies Royo, Eric Mazumdar, Roy Dong, Claire J. Tomlin, S. Shankar Sastry
IEEE Conference on Decision and Control (CDC), 2020 [PDF]

On Approximate Thompson Sampling with Langevin Algorithms
Eric Mazumdar*, Aldo Pacchiano*, Yi-an Ma*, Peter L. Bartlett, Michael I. Jordan
International Conference on Machine Learning (ICML), 2020 [PDF]
(* denotes equal contribution)

Feedback Linearization for Unknown Systems via Reinforcement Learning
Tyler Westenbroek*,David Fridovitch-Keil*, Eric Mazumdar*, Shreyas Arora, Valmik Prabhu, Claire Tomlin, S. Shankar Sastry
International Conference on Robotics and Automation (ICRA), 2020 [PDF]
(* denotes equal contribution)

Policy Gradient Algorithms Have No Guarantees of Convergence in Linear Quadratic Games
Eric Mazumdar, Lillian J. Ratliff, Michael I. Jordan, S. Shankar Sastry
International Conference on Autonomous Agents and Multiagent Systems (AAMAS), 2020 [PDF]

On Gradient-Based Learning in Continuous Games
Eric Mazumdar, Lillian J. Ratliff, S. Shankar Sastry
SIAM Journal on Mathematics for Data Science (SIMODS), 2020 [PDF]

Inverse Risk-Sensitive Reinforcement Learning via Gradient Methods
Lillian Ratliff, Eric Mazumdar
IEEE Transactions on Automatic Control (TAC), 2020 [PDF]

Local Nash Equilibria are Isolated, Strict Local Nash Equilibria in ‘Almost All’ Zero-Sum Continuous Games
Eric Mazumdar and Lillian Ratliff
IEEE Conference on Decision and Control (CDC), 2019 [PDF]

Convergence Analysis of Gradient-Based Learning in Continuous Games
Benjamin, Chasnov, Lillian Ratliff, Eric Mazumdar, Samuel A. Burden
Conference on Uncertainty in Artificial Intelligence (UAI), 2019 [PDF]

On the Analysis of Cyclic Drug Schedules for Cancer Treatment using Switched Dynamical Systems
Margaret Chapman, Eric Mazumdar, Ellen Langer, Rosalie Sears, Claire Tomlin
IEEE Conference on Decision and Control (CDC), 2018 [PDF]

Gradient-based Inverse Risk-Sensitive Reinforcement Learning
Eric Mazumdar, Lillian Ratliff, S. Shankar Sastry
IEEE Conference on Decision and Control (CDC), 2017 [PDF]

To Observe or Not to Observe: Queuing Game Framework for Urban Parking
Lillian Ratliff, Chase Dowling, Eric Mazumdar, Baosen Zhang
IEEE Conference on Decision and Control (CDC), 2016 [PDF]

Understanding the Impact of Parking on Urban Mobility via Routing Games on Queue–Flow Networks
Daniel Calderone, Eric Mazumdar, Lillian Ratliff, S. Shankar Sastry
IEEE Conference on Decision and Control (CDC), 2016 [PDF]

Mathematical Framework for Activity-Based Cancer Biomarkers
Gabriel Kwong, Jaideep Dudani, Emmanuel Carrodeguas, Eric Mazumdar, Miriam Zekvat, Sangeeta N. Bhatia
Proceeding of the National Academy of Science (PNAS), 2015 [PDF]

Workshop Publications

Rethinking Scaling Laws for Learning in Strategic Environments
Tinashe Handina, Eric Mazumdar
Conference on Economics and AI+ML
Econometric Society Interdisciplinary Frontiers (ESIF) [PDF]

Coupled Gradient Flows for Strategic Non-Local Distribution Shift
Lauren Conger, Franca Hoffman, Eric Mazumdar, Lillian Ratliff
L4DC Workshop, International Conference on Machine Learning (ICML), 2023 [PDF]

Policy Gradient Has No Convergence Guarantees in Linear Quadratic Dynamic Games
Eric Mazumdar, Lillian J. Ratliff, Michael I. Jordan, S. Shankar Sastry
Workshop on Smooth Games and Optimization
NeurIPS, 2019 [PDF]

Learning Feedback Linearization by Model-Free Reinforcement Learning
Tyler Westenbroek*, David Fridovitch-Keil*, Eric Mazumdar*, Claire Tomlin, S. Shankar Sastry
Workshop on Generative Modeling/Model-based Reasoning
International Conference on Machine Learning (ICML), 2019
(* denotes equal contribution)

Group

Postdocs:

Zaiwei Chen (co-advised with Adam Wierman)
Laixi Shi (co-advised with Adam Wierman)
Kishan Panaganti Badrinath (co-advised with Adam Wierman)

Students:

Lauren Conger (co-advised with John Doyle)
Tinashe Handina (co-advised with Adam Wierman)
Yizhou Zhang

Prospective Students & Postdocs

Prospective Postdoctoral Fellows

I will be recruiting Postdoctoral fellows in CMS to begin in Fall 2025. More details will be posted here in the coming months, however, if there is an exceptionally good research fit with my group, please feel free to send an email with your CV/resume and a short paragraph highlighting research fit.

Teaching

Caltech Courses

CDS 110 - Analysis and Design of Feedback Control Systems

Fall 2024

This course serves as an introduction to analysis and design of feedback control systems in the time and frequency domain, with an emphasis on state space methods, robustness, and design tradeoffs. Linear input/output systems, including input/output response via convolution, reachability, and observability. State feedback methods, including eigenvalue placement, linear quadratic regulators, and model predictive control. Output feedback including estimators and two-degree of freedom design. Input/output modeling via transfer functions and frequency domain analysis of performance and robustness, including the use of Bode and Nyquist plots. Robustness, tradeoffs and fundamental limits, including the effects of external disturbances and unmodeled dynamics, sensitivity functions, and the Bode integral formula.

CMS/CS/EE/IDS 144 - Network Economics

Winter 2023, 2024

Social networks, the web, and the internet are essential parts of our lives and we all depend on them every day, but do you really know what makes them work? This course studies the "big" ideas behind our networked lives. Things like, what do networks actually look like (and why do they all look seemingly look so similar)? How do search engines work? Why do epidemics spread the way they do? How does computational advertising work? How do we understand how people interact over networks? For all these questions and more, the course will provide a mixture of both mathematical analysis and hands-on projects.

CMS/Ec 248 - Topics in Learning and Games

Fall 2022, 2023

This is an advanced topics course intended for graduate students with a background in optimization, linear systems theory, probability and statistics, and an interest in learning, game theory, and decision making more broadly. We will cover the basics of game theory including equilibrium notions and efficiency, learning algorithms for equilibrium seeking, and discuss connections to optimization, machine learning, and decision theory. While there will be some initial overview of game theory, the focus of the course will be on modern topics in learning as applied to games in non-cooperative settings. We will also discuss games of partial information and stochastic games as well as hierarchical decision-making problems (e.g., incentive and information design).



UC Berkeley Courses

DS 102 - Data, Inference, and Decisions

Spring/Fall 2019

Data Science 102 is a capstone class for the Data Science Major at Berkeley. Its goal is to develop the probabilistic foundations of inference in data science and building a comprehensive view of the modeling and decision-making life cycle in data science including its human, social, and ethical implications.

We covered topics including frequentist and Bayesian decision-making, permutation testing, false discovery rate control, probabilistic interpretations of models, Bayesian hierarchical models, basics of experimental design, confidence intervals, causal inference, multi-armed bandits, Thompson sampling, optimal control, Q-learning, differential privacy, and an introduction to machine learning tools including decision trees, neural networks and ensemble methods.

I participated in the development of this course starting back in Spring 2019, and was a Graduate Student Instructor for the first semester of this course in Fall 2019. This included creating course content (including homeworks, exams, discussions, coding projects, and labs), as well as leading both a discussion and lab section.

IEOR 290 - Data Analytics and the IoT: Machine Learning for Operations with Human Data Sources

Spring 2018

IEOR 290 was an introductory graduate-level class in the Industrial Engineering and Operations Research department on machine learning on data from human data sources and reasoning about the human, social, and ethical implications of making decisions based on such analyses. This course covered the theoretical tools for the analysis of data and human agents in cyber-physical systems. Concepts will included optimization, game theory, differential privacy, behavioral methods, statistical estimation, and utility function learning, with a focus on applications in a variety of Internet of Things systems, such as the energy grid, new transportation services, and database privacy. Throughout, we emphasized the underlying mathematical tools required to understand the current research in each of these fields. I was a Graduate Student Instructor for this class and also participated in the development of this course.