Deep Dive Workshop
Machine Learning Theory
Organized by:
Friday 9 October 10.00
Lead organizers: Nirupam Gupta, Tenure-Track Assistant Professor and Christian Igel, Full Professor, Department of Computer Science, University of Copenhagen
We welcome submissions of ongoing, recently published, or accepted work across all areas of ML theory, targeting MSc, PhD students, and postdocs.
Authors may submit for a long talk (20-30 minutes) or a poster with spotlight presentation, and may indicate willingness for either format. Tentative topics include (but not limited to) design and analysis of learning algorithms, optimization, online and reinforcement learning, deep learning theory, privacy, fairness and robustness.
10:00 – 10:30
Theoretical Foundations of Reliable Learning with Imperfect Data
Arsen Vasilyan, Aarhus University
10:30 – 10:40
Spectral Arm Elimination for Stochastic Bandits with Graph-Structured Feedback
Gholamreza Omidi Ardali, University of Copenhagen
10:40 – 10:50
Structure Predicts Supervision: Minimizing the Length-Generalization Gap in Looped Transformers
Monik Raj Behera, Halmstad University College
10:50 – 11:00
Tail-Freeze Conformalized Quantile Regression
Mikołaj Mazurczyk, University of Copenhagen
11:00 – 11:30
Random Forests under Differential Privacy
Christian Lebeda, University of Copenhagen
Lunch break
13:10 – 13:40
Why can’t we learn privately? Lower bounds for Differentially Private Learning
Amartya Sanyal, University of Copenhagen
13:40 – 13:50
The Symmetries of Three-Layer ReLU Networks
Johanna Marie Gegenfurtner, Technical University of Denmark
13:50 – 14:00
Variance-Aware UCB for Online Task Assignment with Unknown Processing Times and Bandit Feedback
Ahmad Raeisi, University of Tehran
Intermediate: For attendees who have basic understanding or some experience with the subject but are not yet advanced.