
Frederik Kunstner Wins National AI Dissertation Award for Groundbreaking Work on Machine Learning
Recent PhD graduate Frederik Kunstner has been awarded the 2025 Best Doctoral Dissertation Award from the Canadian Artificial Intelligence Association / Association pour l'Intelligence Artificielle au Canada (CAIAC). This prestigious annual award recognizes an outstanding PhD thesis completed at a Canadian university in the field of artificial intelligence. As part of the honour, Frederik presented his research at the recent Canadian AI Conference.
His dissertation, Why Do Machine Learning Optimizers That Work, Work?, explores the underlying mechanisms of commonly used optimization algorithms in machine learning, many of which have been successful in practice but remain poorly understood in theory. His work lays the foundation for more effective algorithm design by identifying why these optimization methods work and how they address the specific challenges of modern machine learning models.

Frederik joined UBC after earning his bachelor's and master's degrees from the École Polytechnique Fédérale de Lausanne (EPFL) in Switzerland, where he worked in the Machine Learning and Optimization Laboratory.
At UBC, he worked under the supervision of Professor Mark Schmidt to better understand and formally explain the effectiveness of widely used optimization heuristics in training complex machine learning models.
In machine learning, optimization algorithms enable models to learn from data by adjusting their own parameters. While optimization is a well-established mathematical field, its theory does not extend well to the complex models used in practice, which often rely on heuristics discovered through trial and error. Frederik's work with Mark aimed to formalize this training process by investigating why these heuristics work, identifying which aspects of the model, data, or training pipeline make optimization challenging. An earlier aspect of this work won Frederik the Best Paper award at AISTATS in 2021.
Frederik is now continuing his research on machine learning optimization as a postdoctoral fellow at INRIA Paris – The National Institute for Research in Digital Science and Technology, funded by a Marie Skłodowska-Curie Postdoctoral Fellowship.
Further Information:
- Frederik's Dissertation: https://open.library.ubc.ca/soa/cIRcle/collections/ubctheses/24/items/1.0445444
- CAIAC award recipients: https://www.caiac.ca/en/best-phd-award
- AISTATS 2021 Best Paper: https://virtual.aistats.org/virtual/2021/poster/1905
- UBC CS 2021 Article: https://www.cs.ubc.ca/news/2021/05/algorithm-research-gets-maximum-results
- Marie Skłodowska-Curie Fellowship: https://marie-sklodowska-curie-actions.ec.europa.eu/actions/postdoctoral-fellowships