Within the last decade, deep neural networks have emerged as an indispensable tool in many areas of artificial intelligence including computer vision, computer graphics, natural language processing, speech recognition and robotics. This reading course will introduce the practical and theoretical principles of deep neural networks with an eye towards their interplay with differential geometry.
We expect the course to have three parts:
basics in statistics, supervised learning and linear regression;
deep learning and convolutional neural network;
topics in geometric deep learning and graph representation learning.
This reading seminar is suitable for Master students, graduate students and postdocs in pure mathematics who want to learn more about deep learning and its recent connections with differential geometry.
We recommend the following books:
T. Hastie, R. Tibshirani, J. Friedman - Elements of Statistical Learning (2nd edition)
Goodfellow, Y. Bengio, A. Courville - Deep Learning
A. Geiger, Lecture notes: Deep Learning - Universitaet Tuebingen (available online)
Bishop - Pattern Recognition and Machine Learning
M. Bronstein, J. Bruna, T. Cohen, P. Velickovic - Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges (Available on arXiv:2104.13478)
The reading course will assume some basic knowledge of probability, linear algebra, and differential geometry.
The seminar course will start on October 27th and will be held weekly on Wednesdays 09:30 - 11:00 in Seminarraum 3 (3rd floor). All the lectures will be taught in English.
Here is the calendar:
27/10/21 - Week 1 : Overview of Supervised Learning I (Chapter 2 : Elements of Statistical Learning) - speaker: Anja Randecker
Week 2: Overview of Supervised Learning II (Chapter 3 : Elements of Statistical Learning) - speaker: Merik Niemeyer
Week 3: Linear Methods for Classification (Chapter 4 : Elements of Statistical Learning) - speaker: Marta Magnani
Week 4: Intro to Neural Networks (Chapter 11: Elements of Statistical Learning ) - speaker: Max Riestenberger
Week 5: Model Assessment and Prediction Errors (Chapter 7: Elements of Statistical Learning) - speaker : Valentina Disarlo
Week 6: Deep Learning I - speaker : Dia Taha
Week 7: Deep Learning II - speaker : Brice Lousteau
Christmas Break
26/01/22: Deep Learning III - speaker : Max Schmahl
2/02/22: Deep Learning IV - speaker: Valentina Disarlo
9/02/22: Deep Learning V - speaker: Dia Taha
16/02/22 : Deep Learning VI - speaker : Brice Lousteau
We are happy to announce the Geometric Deep Learning Hackathon GDLH2022 at Foyer of Marsilius Kolleg (Im Neuenheimer Feld 130.1, 69120 Heidelberg) from Friday 11.02.2022 to Sunday 13.02.2022. The registration form can be found here (the registration is open until 20.01.2022).
Dr. Valentina Disarlo vdisarlo@mathi.uni-heidelberg.de -- Mathematikon -- Room 03.327
Marta Magnani mmagnani@mathi.uni-heidelberg.de -- Mathematikon -- Room 05.314