Mathematics of Deep Learning

Meeting on Mathematics of Deep Learning

Location: TU Delft, Science Centre
Date: 5 November 2019
Organization committee: Remco Duits, Arnold Heemink, Johannes Schmidt-Hieber, Willem Kruijer

Presentations:

Estimation ability of deep learning with connection to sparse estimation in function space
Taiji Suzuki – Department of Mathematical Informatics, University of Tokyo

Learned SVD - Deep Learning Decomposition for Inverse Problems 
Christoph Brune – Department of Applied Mathematics, University of Twente

Functional Process Priors for CNNs and VAEs 
Max Welling, Institute of Informatics, University of Amsterdam

Deep limits of residual neural networks
Yves van Gennip, Delft Institute of Applied Mathematics, Delft University of Technology

Group Equivariant CNNs beyond Roto-Translations:
B-Spline CNNs on Lie Groups
Erik Bekkers, Department of Mathematics and Computer Science, Eindhoven University of Technology

Implicit bias and regularization in machine learning
Lorenzo Rosasco, Laboratory for Computational and Statistical Leaning, Massachusetts Institute of Technology.

Diffusion Variational Autoencoders
Jim Portegies, Department of Mathematics and Computer Science, Eindhoven University of Technology

 

Approximation with sparsely connected deep networks
Remi Gribonval, Centre de Recherche INRIA Rennes

Gauge Equivariant Convolutional Networks
Taco Cohen, Institute of Informatics, University of Amsterdam

 




PDE-based CNNs with Morphological Convolutions
Bart Smets, Department of Mathematics and Computer Science, Eindhoven University of Technology

Photografer: Marc Blommaert