Pascale Gourdeau

Postdoctoral Research Fellow at the Vector Institute

I started as a Postdoctoral Research Fellow at the Vector Institute in October 2023, with a joint appointment at the University of Toronto. I am advised by Nicolas Papernot and Shai Ben-David, and work on the theoretical foundations of trustworthy machine learning.

Previously, I obtained a DPhil (PhD) at the University of Oxford under the supervision of James Worrell, Varun Kanade and Marta Kwiatkowska. My thesis studied the sample complexity of robust learning under evasion (test-time) attacks.

I obtained my masters at McGill University's Reasoning and Learning Lab and was supervised by Prakash Panangaden and Doina Precup. My thesis focused on quantifying the behavioural distance between systems.

For more information, you can have a look at my resume.

Theses

• Sample Complexity of Robust Learning against Evasion Attacks (DPhil thesis)
  P. Gourdeau
  Oxford University Research Archive, 2023.
• Bisimulation Pseudometrics for Weighted Finite Automata (MSc thesis)
  P. Gourdeau
  McGill University Libraries, 2017.

Journal Publications

• On the hardness of robust classification
  P. Gourdeau, V. Kanade, M. Kwiatkowska, and J. Worrell
  In Journal of Machine Learning Research, 2021.
  Paper.
• Bisimulation metrics and norms for real-weighted automata
  B. Balle, P. Gourdeau, and P. Panangaden
  In Information and Computation , 2020.
  Paper.

Conference Publications

For the most up-to-date list, see Google scholar.

• On the Computability of Robust PAC Learning
  P. Gourdeau, T. Lechner, and R. Urner
  In Conference on Learning Theory (COLT) , 2024.
  Paper (arxiv version).
• When are local queries useful for robust learning?
  P. Gourdeau, V. Kanade, M. Kwiatkowska, and J. Worrell
  In 36th Conference on Neural Information Processing Systems (NeurIPS), 2022.
  Paper. Slides.
• Sample complexity bounds for robustly learning decision lists against evasion attacks
  P. Gourdeau, V. Kanade, M. Kwiatkowska, and J. Worrell
  In International Joint Conference in Artifical Intelligence (IJCAI), 2022.
  Paper. Poster. Slides (long presentation).
• On the hardness of robust classification
  P. Gourdeau, V. Kanade, M. Kwiatkowska, and J. Worrell
  In 33rd Conference on Neural Information Processing Systems (NeurIPS), 2019.
  Paper. Poster. Spotlight Slides.
• Bisimulation metrics for weighted automata
  B. Balle, P. Gourdeau, and P. Panangaden
  In 44th International Colloquium on Automata, Languages, and Programming (ICALP 2017), Schloss Dagstuhl-Leibniz-Zentrum fuer Informatik , 2017.
  Paper. Slides.
• Feature selection and oversampling in analysis of clinical data for extubation readiness in extreme preterm infants
  P. Gourdeau, L. Kanbar, W. Shalish, G. Sant'Anna, R. Kearney, and D. Precup
  In 2015 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pp. 4427-4430, IEEE, 2015.
  Paper. Poster.