Amir R. Asadi

Leverhulme Early Career Fellow
Statistical Laboratory,
University of Cambridge

I am a Leverhulme Early Career Fellow at the Statistical Laboratory, part of the Department of Pure Mathematics and Mathematical Statistics at the University of Cambridge, a position I have held since September 2023. I am also a Fellow of the Isaac Newton Trust. From 2021 to 2023, I was as a Postdoctoral Research Associate at the same department. I earned my M.A. and Ph.D. in Electrical and Computer Engineering from Princeton University in 2017 and 2021, respectively, after completing dual B.Sc. degrees in Mathematics and Electrical Engineering from Sharif University of Technology in 2015. My research interests include machine learning, differential privacy, and information theory.

Download CV

Education

B.Sc. in Mathematics

B.Sc. in Electrical Engineering

2010 - 2015

M.A. in Electrical Engineering

2015 - 2017

Ph.D. in Electrical and Computer Engineering

2017 - 2021

Awards and Honours

Leverhulme Early Career Fellowship from the Leverhulme Trust and the Isaac Newton Trust (2023-present)
Teaching Assistant Award from Department of Electrical and Computer Engineering at Princeton University (2019)
Anthony Ephremides Fellowship from Princeton University (2016)
Bronze Medal in the Iranian Mathematical Olympiad (2009)
Diploma of Mathematics, Tournament of Towns Contest, from the Russian Academy of Sciences (2009)
Member of Iran's National Elite Foundation (2009-present)

Publications

A. R. Asadi, A. Davoodi, R. Javadi & F. Parvaresh. (2026) Exact Recovery in the Data Block Model. arXiv preprint arXiv:2602.05852 [Link]
B. Abdolmaleki, A. R. Asadi , V. R. Asadi, S. Köpsell, B. Mohee, N. Roustaeifar & M. Zarezadeh. (2026) VeriDP: Verifiable Differentially Private Training To appear in Privacy Enhancing Technology Systems (PETS).
G. Aminian, I. Shenfeld, A. R. Asadi, A. Beirami & Y. Mroueh. (2026) Best-of-N through the Smoothing Lens: KL Divergence and Regret Analysis. To appear in International Conference on Learning Representations (ICLR). [Link]
A. R. Asadi. (2025) Hierarchical Maximum Entropy via the Renormalization Group. arXiv preprint arXiv:2509.01424 (Submitted) [Link]
G. Aminian, A. R. Asadi, I. Shenfeld & Y. Mroueh. (2025) Theoretical Analysis of KL-regularized RLHF with Multiple Reference Models. Advances in Neural Information Processing Systems (NeurIPS). [Link]
G. Aminian, A. R. Asadi, T. Li, A. Beirami, G. Reinert & S. N. Cohen. (2025) Generalization and Robustness of the Tilted Empirical Risk. International Conference on Machine Learning (ICML). [Link]
A. Pensia, A. R. Asadi, V. Jog & P.-L. Loh. (2024) Simple Binary Hypothesis Testing under Local Differential Privacy and Communication Constraints. IEEE Transactions on Information Theory vol. 71, no. 1, pp. 592-617 [Link]
A. R. Asadi. (2024) An Entropy-Based Model for Hierarchical Learning. Journal of Machine Learning Research, 25(187), pp. 1−45. [Link]
A. R. Asadi & P.-L. Loh. (2024) Entropic Regularization of Neural Networks: Self-Similar Approximations. Journal of Statistical Planning and Inference, Vol. 233, [Link]
A. R. Asadi & P.-L. Loh. (2023) On the Gibbs Exponential Mechanism and Private Data Generation. IEEE International Symposium on Information Theory (ISIT) pp. 2213-2218, [Link]
A. Pensia, A. R. Asadi, V. Jog & P.-L. Loh. (2023) Simple Binary Hypothesis Testing under Local Differential Privacy and Communication Constraints. Conference on Learning Theory (COLT). [Link]
A. R. Asadi & E. Abbe. (2020) Chaining Meets Chain Rule: Multilevel Entropic Regularization and Training of Neural Networks. Journal of Machine Learning Research, 21(139), pp. 1-32. [Link]
A. R. Asadi, E. Abbe, & S. Verdú. (2018) Chaining Mutual Information and Tightening Generalization Bounds. Advances in Neural Information Processing Systems (NeurIPS), pp. 7245-7254. [Link]
A. R. Asadi, E. Abbe, & S. Verdú. (2017) Compressing Data on Graphs with Clusters. IEEE International Symposium on Information Theory (ISIT) pp. 1583-1587. [Link]
Majid Asadi, A. R. Asadi. (2014) On the Failure Probability of Used Coherent Systems. Communications in Statistics, Theory and Methods, Vol. 43, pp. 2468-2475. [Link]

Ph.D. Dissertation

A. R. Asadi (2021) Neural Network Learning: A Multiscale-Entropy and Self-Similarity Approach. (Princeton University) [Link]

Google Scholar

Talks

Informed-AI Hub, United Kingdom. November 2025
UK Crypto Day, University of Sheffield, Sheffield, United Kingdom. June 2025
Chennai Mathematical Institute, Chennai, India. January 2025
Department of Mathematical Sciences, Durham University, UK. May 2023
Department of Computer Science, ETH Zürich, Switzerland. February 2021
NSF-Simons Collaboration on the Theoretical Foundations of Deep Learning. December 2020
Department of EECS, Massachusetts Institute of Technology, USA. December 2020
Center for Data Science, New York University, USA. June 2020
Laboratoire de Physique, École Normale Supérieure, France. May 2020
Department of Statistical Sciences, University of Toronto, Canada. April 2020
Institute for Advanced Study, Princeton, USA. October 2019 [YouTube Link]
Microsoft Research AI, Redmond, USA. September 2019

Workshop Organization

Co-organizer, 1st Cambridge Information Theory Colloquium, 21 April 2023
Co-organizer, 2nd Cambridge Information Theory Colloquium, 10 May 2024
Co-organizer, 8th London Symposium on Information Theory, 14-16 May 2025

Personal

Outside my academic work, I enjoy exercise and reading poetry, particularly the works of Saadi and Hafez.