• This module provides students with both theoretical and practical insights into modern Deep Learning
• In particular, we focus on a novel approach for understanding deep neural networks with mathematical tools from geometry and group theory
• This enables a methodical approach to Deep Learning: starting from first principles of symmetry and invariance, we derive different network architectures for analyzing unstructured sets, grids, graphs, and manifolds
• Topics of the course include: group theory, graph neural networks, convolutional neural networks, applications of geometric deep learning in diverse fields such as geometry processing, molecular dynamics, social networks, game playing (computer Go), processing of text and speech, as well as applications in medicine

M. M. Bronstein, J. Bruna, T. Cohen, P. Veličković. Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges
https://arxiv.org/pdf/2104.13478.pdf

Kevin P. Murphy. Machine Learning: A Probabilistic Perspective.
MIT Press, 2012

I. Goodfellow, Y. Bengio, A. Courville. Deep Learning.
MIT Press, 2017

Further material on https://geometricdeeplearning.com

Informationen zu mündlichen Prüfungen - im ILIAS Portal