ELISE Theory Workshop on Machine Learning Fundamentals
5th - 7th September 2022
EURECOM, Sophia Antipolis, France
About
This workshop aims at encouraging communication between researchers working on fundamental aspects of machine learning. Topics include, but are not limited to, Gaussian processes, kernel methods, probabilistic numerics, interpretability, optimization, and deep learning.
Schedule
September 5 (Mon)
10:00 - 10:10: Opening
10:10 - 11:00 Talk 1: Alix Lheritier (Amadeus)
Sequential Nonparametric Two-Sample Testing via Universal Prediction [slides]
11:10 - 12:00 Talk 2: Nicholas Krämer (University of Tuebingen)
Probabilistic numerical simulation of partial differential equations
12:00 - 14:00 Lunch break
14:00 - 14:50 Talk 3: Maurizio Filippone (EURECOM)
Functional Priors for Bayesian Deep Learning [slides]
15:00 - 16:30 Poster session
September 6 (Tue)
10:00 - 10:50 Talk 1: Antonin Schrab (University College London)
Aggregated Kernel Tests [slides]
11:00 - 11:50 Talk 2: Junhyung Park (Max Planck Institute)
Kernel conditional mean embeddings, and theory of functional response regression [slides]
12:00 - 14:00 Lunch break
14:00 - 14:50 Talk 3: Luc Pronzato (CNRS)
Kernel relaxation for space-filling design [slides]
15:00 - 15:50 Talk 4: Chau Siu Lun (University of Oxford)
Explainability for Kernel Methods [slides]
16:00 - 16:50 Talk 5: Damien Garreau (Université Côte d'Azur & INRIA)
What does LIME really see in images?
September 7 (Wed)
10:00 - 10:50 Talk 1: Michael Arbel (INRIA)
Bilevel Optimization in Machine Learning: Beyond Strong Convexity [slides]
11:00 - 11:50 Talk 2: Toni Karvonen (University of Helsinki)
Analytic kernels in Gaussian process interpolation [slildes]
12:00 - 14:00 Lunch break
14:00 - 14:50 Talk 3: Felix Dangel (University of Tuebingen)
Backpropagation Beyond the Gradient [slides]
15:00 - 15:50 Talk 4: Pierre-Alexandre Mattei (INRIA)
Safe semi-supervised learning [slides]
16:00 - 17:00 Discussion
Poster session (Sep 5th, Mon, 3 pm)
Julia Grosse (Univeristy of Tuebingen)
Optimistic Optimization of Gaussian Process Samples
Heishiro Kanagawa (University College London)
When does kernel Stein discrepancy detect (non)convergence of moments?
Simone Rossi (EURECOM)
Model Selection for Bayesian Autoencoders
Jonas Wacker (EURECOM)
Local Random Feature Approximations of the Gaussian Kernel
Giulio Franzese (EURECOM)
A New Look on Diffusion Times for Score-based Generative Models
Louis Ohl (Université Côte d'Azur & INRIA)
Discriminative Clustering using Distance-Generalised Mutual Information
Hugo Schmutz (Université Côte d'Azur & INRIA)
Don't fear the unlabelled: Safe semi-supervised learning via simple debiasing
Tamim El Ahmad (Telecom Paris)
p-Sparsified Sketches for Fast Multiple Output Kernel Methods
Organizers
Motonobu Kanagawa (EURECOM)
Arthur Gretton (University College London)
Philipp Hennig (University of Tuebingen & Max Planck Institute for Intelligent Sytems)
Venue
Amphitheater (level 0), EURECOM, Campus SophiaTech, 450 Route des Chappes, 06410 Biot [Google Map]
Registration
If you are interested in participating in the workshop, please register yourself here.
Contact
If you have any questions, please send an email to motonobu.kanagawa [RKHS] eurecom.fr (Please replace [RKHS] by @).