y [dot] jedra [at] ic [dot] ac [dot] uk | Google Scholar
I am joining Imperial College London as assistant professor within the EEE department starting 2025/2026!
As of September 2025, I am an assistant professor in EEE Department of Imperial College London.
My research interests broadly revolve around developing statistical and algorithmic foundations for sequential learning and control of dynamical systems under uncertainty. In particular, I have been working on interconnected topics including reinforcement learning, multi-armed bandits, system identification, adaptive control, & high-dimensional/causal inference with the aim of understanding how to exploit structural characterisations -- of practical relevance in societal, scientific and engineering domains -- to enable fast and efficient learning.
Previously I was a postdoctoral researcher in LIDS at MIT working with Devavrat Shah, where I was generously supported by the Knut and Alice Wallenberg Foundation. I completed my PhD in Electrical Engineering from KTH Royal Institute of Technology, where I was advised by Alexandre Proutiere. During my PhD, I was also part of the WASP AI program. Before that, I received a MSc degree in Applied and Computational Mathematics within the Engineering Physics' program at KTH and both my BSc degree and MSc degree in Mathematics and Computer Science from ENSIMAG -- Grenoble-INP, in 2015 and 2018, respectively.
[New!] "Model-free Low-rank RL with Leveraged Entrywise Matrix Estimation" joint work with S. Stojanovic and A. Proutiere, previously accepted at NeurIPS 2024 will be presented at EWRL 2025! (arXiv:2506.08720)
[New!] "Minimal Order Recovery through Rank-adaptive Identification" joint work with F. Zhang and A. Proutiere, is now accepted at CDC 2025! (arXiv:2506.08720)
[New!] "Sub-optimality of the Separation Principle for Quadratic Control from Bilinear Observations" joint work with S. Choi, Y. Sattar, M. Fazel and S. Dean, is now accepted at CDC 2025! (arXiv:2504.11555)
[New!] "Finite Sample Identification of Partially Observed Bilinear Dynamical Systems" joint work with Y. Sattar, M. Fazel and S. Dean is now accepted at L4DC 2025!( arXiv:2501.07652)
-> Happy to announce that it has been also selected for an oral presentation!
[07/2025] "Optimal Transfer Learning For Missing-Not-At-Random Matrix Completion" joint work with A. Jalan, A. Mazumdar, S. Mukherjee and P. Sarkar is now on arXiv:2503.00174
[07/2025] "Learning Linear Systems from Bilinear measurements" joint work with Y. Sattar and S. Dean was presented at ACC 2025! (arXiv:2409.16499)
[02/2025] I am hosted by A. Rantzer to serve as opponent to the Licentiate Thesis of F. Bencherki on "Adaptive Control of Positive Systems".
[02/2025] "k-SVD with Gradient Descent" joint work with E. Gan and D. Shah is now on arXiv! (arXiv:2502.00320). We offer a purely gradient based method for computing SVD!
[12/2024] "Model-free Low-rank RL with Leveraged Entrywise Matrix Estimation" was accepted at Neurips 2024 in Vancouver
[10/2024] Presenting a talk on "Exploiting observation bias to improve matrix completion" at Informs 2024 in Seattle
[08/2024] Attending 2024 ESIF Economics and AI+ML meeting! I will be presenting our recent work with Devavrat on causal inference, more specifically on how to exploit observation bias to improve outcome prediction
[07/2024] Attending ICML 2024! We will be presenting our work on low-rank bandits using tight two-to-inifty subspace recovery bounds
[07/2024] Attending Stochastic Networks 2024! Stefan and William presented posters on some joint work on Low-rank in bandits and RL
[12/2023] Attending Neurips 2023!
[12/2023] Ingvar, Anastasios and Bruce are presenting our tutorial on Non-asymptotic Theory of SysID at CDC2023
[09/2023] Starting Fall 2023 I will be a Postdoctoral Fellow at LIDS, MIT working with Prof. Devavrat Shah.
[09/2023] Our work on Entrywise Guarantees for Matrix Estimation in RL has been accepted to Neurips 2023
[09/2023] Our work on discounted linear MDPs has been accepted at EWRL 2023
[06/2023] Attending NORDSTAT 2023 where I presented our work on BlockMDPs
[06/2023] I successfully defended my PhD thesis titled "Statistical Learning in Linearly Structured Systems: Identification, Control, and Reinforcement Learning". I am grateful for having Prof. Alexandre Rakhlin as my opponent, and Rs. Sci. Alessandro Lazaric, Ass. Prof. Sarah Dean, and Prof. Csaba Szepesvári as members of my thesis committee
[04/2023] Attending AISTATS 2023 in Valencia to present our work on Block MDPs together with Junghyun Lee
[03/2023] Presenting our work on Block MDPs at 2nd ASCAI workshop in TUM
[12/2022] Presenting our work on regret minimization for LQR at the IEEE CDC 2022 full-day workshop on Statistical Learning Theory for Control
[09/2022] Atteneding the Simons Institute Workshop on Quantifying Uncertainty: Stochastic, Adversarial, and Beyond.
Below is a selected list of my publications. Please check my research page and Google Scholar for an updated list of my publications.
(α - β) means that authorship is in alphabetical order.
Exploiting Observation Bias to Improve Matrix Completion. | Causal < Stats |
(α - β) : Yassir Jedra, Sean Mann, Charlotte Park, Devavrat Shah, (in submission)
Model-free Reinforcement Learning via Leveraged Entry-wise Matrix Estimation | Stats | MDPs < RL |
Stefan Stojanovic, Yassir Jedra, Alexandre Proutiere, NeurIPS 2024
Spectral Entry-wise Matrix Estimation in Low-rank Reinforcement Learning | Stats | Bandits, MDPs < RL|
Stefan Stojanovic, Yassir Jedra, Alexandre Proutiere, NeurIPS 2023
Nearly Optimal Latent State Decoding in Block MDPs | Clustering < Stats | MDPs < RL |
(α - β) : Yassir Jedra, Junghyun Lee, Alexandre Proutiere, Seyoung Yun, AISTATS 2023
Minimal Expected Regret in Linear Quadratic Systems | Control | RL |
Yassir Jedra, Alexandre Proutiere, AISTATS 2022
Finite time-identification of linear system identification: fundamental limits and optimal algorithms | SysId < Control | RL |
Yassir Jedra, Alexandre Proutiere, IEEE TAC 2022
Optimal best-arm identification in linear bandits | Bandits < RL |
Yassir Jedra, Alexandre Proutiere, NeurIPS 2020
Optimal Algorithms for Multiplayer Multi-armed Bandits | Bandits < RL |
Po-an Wang, Alexandre Proutiere, Kaito Ariu , Yassir Jedra, Alessio Russo , AISTATS 2020