Hassan Ashtiani


Assistant Professor
Department of Computing and Software
Faculty of Engineering
McMaster University

Email: zokaeiam@mcmaster.ca
Phone: +1-905-525-9140 ext. 27234
Office: ITB 246
Photo
[Publications] [Bio] [Students]

Research

Broadly speaking, my research interests revolve around Machine Learning, Artificial Intelligence, Statistics, and Theoretical Computer Science. I am interested in formulating new/emerging learning scenarios (including various forms of unsupervised learning), and providing provably efficient methods for -- or establishing inherent limitations in -- solving them.

Some of the directions that I am currently focusing on are
  • Statistically/Computationally Efficient Distribution Learning
  • Differentially Private Distribution Learning
  • Domain Adaptation and Robust Learning under Distribution Shift
  • Learning under Adversarial Attacks
  • Modern Generalization Bounds for Supervised Learning

  • Highlighted Publications [Full List]

    1. Mixtures of Gaussians are Privately Learnable with a Polynomial Number of Samples [paper]
      Mohammad Afzali, Hassan Ashtiani, Christopher Liaw
      Arxiv Preprint

    2. On the Role of Noise in the Sample Complexity of Learning Recurrent Neural Networks: Exponential Gaps for Long Sequences [paper]
      Alireza Fathollah Pour, Hassan Ashtiani
      NeurIPS 2023

    3. Polynomial time and private learning of unbounded Gaussian Mixture Models [paper]
      Jamil Arbas, Hassan Ashtiani, Christopher Liaw
      ICML 2023

    4. Adversarially Robust Learning with Tolerance [paper]
      Hassan Ashtiani, Vinayak Pathak, Ruth Urner
      ALT 2023

    5. Benefits of Additive Noise in Composing Classes with Bounded Capacity [paper]
      Alireza Fathollah Pour, Hassan Ashtiani
      NeurIPS 2022

    6. Private and polynomial time algorithms for learning Gaussians and beyond [paper]
      Hassan Ashtiani, Christopher Liaw
      COLT 2022

    7. On the Sample Complexity of Privately Learning Unbounded High-Dimensional Gaussians [paper]
      Ishaq Aden-Ali, Hassan Ashtiani, Gautam Kamath
      ALT 2021

    8. Near-optimal Sample Complexity Bounds for Robust Learning of Gaussian Mixtures via Compression Schemes [paper], [arXiv]
      Hassan Ashtiani, Shai Ben-David, Nick Harvey, Chris Liaw, Abbas Mehrabian, Yaniv Plan
      Journal of the ACM, 2020

    9. Nearly tight sample complexity bounds for learning mixtures of Gaussians via sample compression schemes [paper]
      Hassan Ashtiani, Shai Ben-David, Nick Harvey, Chris Liaw, Abbas Mehrabian, Yaniv Plan
      NeurIPS (NIPS) 2018, Oral Presentation (Best Paper Award)