Aritra Mitra
 |
Postdoctoral Researcher E-mail: |
About me
I am a Postdoctoral Researcher in the Department of Electrical and Systems Engineering, University of Pennsylvania, where I am working with Professor George Pappas and Professor Hamed Hassani. I received my Ph.D. degree in August 2020 from the School of Electrical and Computer Engineering, Purdue University, where I was advised by Professor Shreyas Sundaram.
Prior to joining Purdue, I received my M.Tech degree from the Indian Institute of Technology, Kanpur in 2015, and my B.E. degree from Jadavpur University, Kolkata in 2013, both in Electrical Engineering.
I will be joining the Department of Electrical and Computer Engineering at North Carolina State University as an Assistant Professor from January 2023. I am looking for motivated PhD students to work on theoretical problems related to control, optimization, and sequential decision-making under uncertainty (e.g., bandits and reinforcement learning), with a particular focus on multi-agent systems. If you are interested in working with me, please feel free to send me an email.
Research Interests
The broad goal of my research is to enable reliable and efficient learning and decision-making in large-scale distributed systems, while contending with modern challenges related to computation, communication, and adversarial robustness. To meet this goal, my research draws on ideas and tools from Control and Optimization Theory, Statistical Signal Processing, Machine Learning, and Network Science. While my work is theoretically grounded, the theory that I develop is motivated by a variety of application domains: multi-robot systems, wireless sensor networks, federated learning, edge-computing, estimation and control in smart cities and power-grids, and learning in social networks.
My recent postdoctoral work focuses on two main themes: (i) Designing fast and communication-efficient algorithms for the emerging paradigm of Federated Learning; and (ii) Investigating the performance bounds of sequential decision-making problems (e.g., bandits and reinforcement learning) in multi-agent settings. Prior to that, my dissertation made fundamental algorithmic and theoretical contributions to the study of state estimation and statistical inference over networks, subject to worst-case adversarial attacks on certain components. A list of keywords that succinctly describe my past and current research interests is as follows.
Multi-Agent Reinforcement Learning and Bandits
Optimization and Statistical Inference
Federated Learning
Learning, Control, and Estimation over Networks
Resilience and Security
Recent Updates
September 2022: Presented a poster on our robust collaborative bandits work at the Quantifying Uncertainty: Stochastic, Adversarial, and Beyond Workshop, Simons Institute for the Theory of Computing.
September 2022: Our paper titled Collaborative Linear Bandits with Adversarial Agents: Near-Optimal Regret Bounds is accepted to NeurIPS 22, New Orleans, USA.
July 2022: Couple of papers accepted for presentation at the Decision and Control Conference (CDC), Cancun, Mexico.
March 2022: Invited Talk at the Department of Electrical and Computer Engineering at North Carolina State University.
March 2022: Invited Talk at the Department of Aerospace Engineering and Engineering Mechanics at the University of Texas, Austin.
March 2022: Invited Talk at the Department of Electrical and Computer Engineering at the University of Michigan, Ann Arbor.
February 2022: Invited Talk at the Department of Electrical and Computer Engineering at Georgia Tech.
February 2022: Invited (Virtual) Talk at the Department of Electrical and Computer Engineering at Rutgers University.
November 2021: Presented our work at Google's Workshop on Federated Learning and Analytics. [ Talk ]
October 2021: Our paper titled Distributed State Estimation Over Time-Varying Graphs: Exploiting the Age-of-Information is accepted to the IEEE Transactions on Automatic Control!
October 2021: Selected to receive a NeurIPS 2021 Outstanding Reviewer Award given to the top 8% of reviewers who were judged to be instrumental to the review process!
September 2021: Our paper titled Linear Convergence in Federated Learning: Tackling Client Heterogeneity and Sparse Gradients is accepted to NeurIPS 21!
September 2021: I gave an invited talk at Google on the Impacts of Statistical Heterogeneity in Federated Supervised Learning and Best-Arm Identification.
September 2021: Our paper titled On the Computational Complexity of the Secure State-Reconstruction Problem is accepted to Automatica.
July 2021: Our papers titled Online Federated Learning and Federated Learning with Incrementally Aggregated Gradients are accepted for presentation at the Decision and Control Conference (CDC), 2021, Austin, Texas, USA.
June 2021: Our paper Distributed Inference with Sparse and Quantized Communication is accepted to the IEEE Transactions on Signal Processing.
March 2021: Our paper Near-Optimal Data Source Selection for Bayesian Learning is accepted to the 3rd Annual Learning for Dynamics and Control Conference (L4DC) 2021, ETH Zurich, Switzerland.
September 2020: Our paper titled A New Approach to Distributed Hypothesis Testing and Non-Bayesian Learning: Improved Learning Rate and Byzantine-Resilience is accepted to the IEEE Transactions on Automatic Control!
July 2020: Our paper titled Event-Triggered Distributed Inference is accepted for presentation at the Decision and Control Conference (CDC) 2020, Jeju Island, Republic of Korea. An extended version of the paper is available here.
July 2020: Gave an invited (virtual) talk on my PhD research at the Indian Institute of Technology, Kharagpur.
June 2020: Successfully defended my PhD thesis!
Jan 2020: Presented our work on Distributed Hypothesis Testing and Non-Bayesian Social Learning at the Robert Bosch Centre for Cyber-Physical Systems, IISC Bangalore. Here is an abstract of the talk.