Randomized Quantization with Exact Error Distribution
Mahmoud Hegazy and Cheuk Ting Li. "Randomized Quantization with Exact Error Distribution." 2022 IEEE Information Theory Workshop (ITW), pages 350-355, 2022.
CV
Download
Education
- B.S. in Mathematics and Information Engineering(MIE), CUHK, 2018-2022
- Graduated from the ELITE stream with First Honours
- PhD-Track in Data and AI (Mathematics), IP-Paris
- M1 in Data and AI, 2023
- M2 Mathématiques, Vision, Apprentissage (MVA)
Work experience
- Summer 2023-now: Student Researcher, CMAP, Polytechnique
Under the supervision of Prof. Aymeric Dieuleveut, I am working on research projects related to federated optimization. In particular, this work tackles methods of leverag new compression schemes to achieve differential privacy, Langevin dynamics, and smoothing. In addition, I am also working on new algorithms extending the gradient-based aggregation schemes used in federated optimization to optimization problems that are no inherentely gradient-based such as EM algorithms.
- Summer 2021: Research Intern, Information Engineering Dept, CUHK
Working with Prof. Cheuk Ting Li, this project aimed to develop channel synthesis schemes for additive noise channels under finite blocklength. The current focus is analyzing the relationship between the achievable bit rate and the additive noise distribution. A paper has been published at the Information Theory Workshop Conference.
- Summer 2020: Research Intern, MLO, EPFL
Under the supervision of Prof. Martin Jaggi, Dr Mary-Anne Hartley, and Dr Sai Karimireddy, I participated in machine learning research endeavours to explore methods of personalized federated learning. This project leveraged a multi-task learning viewpoint of constructing a similarity matrix between clients which is then used to infer clients’ relations for gradient sharing.
Publications
Compression with Exact Error Distribution for Federated Learning
Mahmoud Hegazy, Remi Leluc, Cheuk Ting Li, and Aymeric Dieuleveut. "Compression with Exact Error Distribution for Federated Learning." Proceedings of the International Conference on Artificial Intelligence and Statistics (AISTATS), 2024.
Scalable Utility-Aware Multiclass Calibration
Mahmoud Hegazy, Michael I. Jordan, and Aymeric Dieuleveut. "Scalable Utility-Aware Multiclass Calibration." Under submission, 2025.
Federated Majorize-Minimization: Beyond Parameter Aggregation
Aymeric Dieuleveut, Gersende Fort, Mahmoud Hegazy, and Hoi-To Wai. "Federated Majorize-Minimization: Beyond Parameter Aggregation." arXiv preprint arXiv:2507.17534, 2025.
Neural Optimal Transport Meets Multivariate Conformal Prediction
Vladimir Kondratyev, Alexander Fishkov, Nikita Kotelevskii, Mahmoud Hegazy, Remi Flamary, Maxim Panov, and Eric Moulines. "Neural Optimal Transport Meets Multivariate Conformal Prediction." arXiv preprint arXiv:2509.25444, 2025.
Valid Selection among Conformal Sets
Mahmoud Hegazy, Liviu Aolaritei, Aymeric Dieuleveut, and Michael I. Jordan. "Valid Selection among Conformal Sets." Advances in Neural Information Processing Systems (NeurIPS), 2025.