Thesis

• Robust Accelerated Gradient Methods for Machine Learning
  Alireza Fallah
  Master of Science in Electrical Engineering and Computer Science, Massachusetts Institute of Technology, June 2019.
  (Ernst A. Guillemin Best Thesis Award in Electrical Engineering (1st prize))

Publications

• Bridging Central and Local Differential Privacy in Data Acquisition Mechanisms
  with Ali Makhdoumi, Azarakhsh Malekian, and Asuman Ozdaglar
  Advances in Neural Information Processing Systems (NeurIPS), 2022 

• Optimal and Differentially Private Data Acquisition: Central and Local Mechanisms
  with Ali Makhdoumi, Azarakhsh Malekian, and Asuman Ozdaglar
  – Minor revision, Operations Research, 2022.
  – Accepted to ACM Conference on Economics and Computation (EC), 2022 (extended abstract) [Video].
  – Accepted to ICML Workshop on Theory and Practice of Differential Privacy (TPDP), 2022.

• Robust Distributed Accelerated Stochastic Gradient Methods for Multi-Agent Networks
  with Mert Gürbüzbalaban, Asuman Ozdaglar, Umut Simsekli, and Lingjiong Zhu.
  To appear in Journal of Machine Learning Research (JMLR), 2022.

• Generalization of Model-Agnostic Meta-Learning Algorithms: Recurring and Unseen Tasks
  with Aryan Mokhtari and Asuman Ozdaglar
  Advances in Neural Information Processing Systems (NeurIPS), 2021.

• On the Convergence Theory of Debiased Model-Agnostic Meta-Reinforcement Learning
  with Kristian Georgiev, Aryan Mokhtari, and Asuman Ozdaglar
  Advances in Neural Information Processing Systems (NeurIPS), 2021.

• Entropic Compressibility of Lévy Processes   
  with Julien Fageot and Thibaut Horel.
  IEEE Transactions on Information Theory, 2022.

• Private Adaptive Gradient Methods for Convex Optimization
  with Hilal Asi, John Duchi, Omid Javidbakht, and Kunal Talwar.
  International Conference in Machine Learning (ICML), 2021.

• A Wasserstein Minimax Framework for Mixed Linear Regression
  with Theo Diamandis, Yonina Eldar, Farzan Farnia, and Asuman Ozdaglar.
  International Conference in Machine Learning (ICML), 2021.

• Personalized Federated Learning with Theoretical Guarantees: A Model-Agnostic Meta-Learning Approach 
  with Aryan Mokhtari and Asuman Ozdaglar.
  Advances in Neural Information Processing Systems (NeurIPS), 2020.

• On the Convergence Theory of Gradient-Based Model-Agnostic Meta-Learning Algorithms
  with Aryan Mokhtari and Asuman Ozdaglar.
  International Conference on Artificial Intelligence and Statistics (AISTATS), 2020.

• An Optimal Multistage Stochastic Gradient Method for Minimax Problems
  with Asuman Ozdaglar and Sarath Pattathil.
  IEEE Conference on Decision and Control (CDC), 2020.

• A Universally Optimal Multistage Accelerated Stochastic Gradient Method
  with Necdet Serhat Aybat, Mert Gürbüzbalaban, and Asuman Ozdaglar.
  Advances in Neural Information Processing Systems (NeurIPS), 2019.

• Robust Accelerated Gradient Methods for Smooth Strongly Convex Functions
  with Necdet Serhat Aybat, Mert Gürbüzbalaban, and Asuman Ozdaglar.
  SIAM Journal on Optimization (SIOPT), Volume 30, Issue 1, pages 717-751, 2020.

• Multidimensional Lévy White Noise in Weighted Besov Spaces
  with Julien Fageot and Michael Unser.
  Stochastic Processes and their Applications, Volume 127, Issue 5, 2017.

• Sampling and Distortion Tradeoffs for Indirect Source Retrieval
  with Elahe Mohammadi and Farokh Marvasti.
  – IEEE Transactions on Information Theory, vol. 63, no. 11, pp. 6833-6848, 2017.
  – 2016 IEEE Global Conference on Signal and Information Processing (GlobalSIP).