Hilal Asi

I am a researcher in the Machine Learning Research (MLR) team at Apple where I work primarily on privacy-preserving machine learning. I obtained my PhD from Stanford University, where I was advised by John Duchi. Previously, I obtained a B.Sc. and M.Sc. from the department of computer science at the Technion, where I was advised by Eitan Yaakobi. During my PhD, I also spent some time at Apple where I worked with Vitaly Feldman and Kunal Talwar.

My current research interests include privacy-preservin machine learning and its intersection with other fields such as optimization and robustness.

Email: (first name).(last name) 94 @ gmail.com

Preprints

• Private Online Leaning via Lazy Algorithms
  with Tomer Koren, Daogao Liu, Kunal Talwar. [pdf]
• DP-Dueling: Learning from Preference Feedback without Compromising User Privacy
  with Aadirupa Saha. [pdf]
• Near Instance-Optimality in Differential Privacy
  with John Duchi. [pdf]

Conference publications

• Universally Instance-Optimal Mechanisms for Private Statistical Estimation
  with John Duchi, Saminul Haque, Zewei Li, Feng Ruan.
  COLT, 2024. [pdf]
• Private Vector Mean Estimation in the Shuffle Model: Optimal Rates Require Many Messages
  with Vitaly Feldman, Jelani Nelson, Huy L Nguyen, Samson Zhou, Kunal Talwar.
  ICML, 2024. [pdf]
• User-level differentially private stochastic convex optimization: Efficient algorithms with optimal rates
  with Daogao Liu.
  AISTATS, 2024. [pdf]
• Faster optimal LDP mean estimation via random projections
  with Vitaly Feldman, Jelani Nelson, Huy L. Nguyen, Kunal Talwar.
  NeurIPS, 2023. [pdf]
• Near-Optimal Algorithms for Private Online Optimization in the Realizable Regime
  with Vitaly Feldman, Tomer Koren, Kunal Talwar.
  ICML, 2023. [pdf]
• From Robustness to Privacy and Back
  with Jonathan Ullman, Lydia Zakynthinou.
  ICML, 2023. [pdf]
• Private Online Prediction from Experts: Separations and Faster Rates
  with Vitaly Feldman, Tomer Koren, Kunal Talwar.
  COLT, 2023. [pdf]
• Optimal Algorithms for Mean Estimation under Local Differential Privacy
  with Vitaly Feldman, Kunal Talwar.
  ICML, 2022 (oral presentation). [pdf]
• Private optimization in the interpolation regime: faster rates and hardness results
  with Karan Chadha, Gary Cheng, John Duchi.
  ICML, 2022.
• Element Level Differential Privacy
  with John Duchi, Omid Javidbakht. [pdf]
  PPAI, 2022.
• Stochastic Bias-Reduced Gradient Methods
  with Yair Carmon, Arun Jambulapati, Yujia Jin, Aaron Sidford. [pdf]
  NeurIPS, 2021.
• Adapting to function difficulty and growth conditions in private optimization
  with Daniel Levy, John Duchi.
  NeurIPS, 2021.
• Private Stochastic Convex Optimization: Optimal Rates in L1 Geometry
  with Vitaly Feldman, Tomer Koren, Kunal Talwar.
  ICML, 2021 (oral presentation). [pdf]
• Private Adaptive Gradient Methods for Convex Optimization
  with John Duchi, Alireza Fallah, Omid Javidbakht, Kunal Talwar.
  ICML, 2021. [pdf]
• Instance-optimality in differential privacy via approximate inverse sensitivity mechanisms
  with John Duchi.
  NeurIPS, 2020. [pdf]
• Minibatch Stochastic Approximate Proximal Point Methods
  with Karan Chadha, Gary Cheng, John Duchi.
  NeurIPS, 2020 (spotlight). [pdf]
• Modeling simple structures and geometry for better stochastic optimization algorithms
  with John Duchi.
  AISTATS, 2019. [pdf]
• Nearly Optimal Constructions of PIR and Batch Codes
  with Eitan Yaakobi.
  ISIT, 2017. [pdf]

Journal publications

• The importance of better models in stochastic optimization
  with John Duchi.
  Proceedings of the National Academy of Sciences, 2019. [pdf] [code]
• Stochastic (Approximate) Proximal Point Methods: Convergence, Optimality, and Adaptivity
  with John Duchi.
  SIAM Journal on Optimization, 2019. [pdf]
• Nearly Optimal Constructions of PIR and Batch Codes
  with Eitan Yaakobi.
  IEEE Transactions on Information Theory, 2019. [pdf]

Teaching

• Fall 2021 Teaching assistant for Stats311: Information Theory and Statistics taught by John Duchi.
• Fall 2019 Teaching assistant for EE364A: Convex Optimization taught by John Duchi.

Service

• Reviewer: ICML, NeurIPS (outstanding reviewer award 2021), AISTATS, SIAM Optimization, JMLR, IEEE Information theory