Beyond the Patterns 36 - Gitta Kutyniok – Deep Neural Networks: The Mystery of Generalization

We are very proud to welcome Gitta Kutyniok from LMU Munich to our lab! Abstract: One or maybe the main reason for the impressive success of deep neural networks in both public life and science is their amazing generalization ability, namely their performance on unseen data. However, this phenomenon is still to a large extent a mystery. In this talk, we will provide an introduction to this problem and discuss some recent advances. We will then focus on graph convolutional neural networks and show how to unravel part of the mystery in this situation completely. Short Bio: Kutyniok was educated in Detmold, and in 1996 earned a diploma in mathematics and computer science at Paderborn University. She completed her doctorate (Dr. rer. nat.) at Paderborn in 2000. Her dissertation, Time-Frequency Analysis on Locally Compact Groups, was supervised by Eberhard Kaniuth. From 2000 to 2008 she held short-term positions at Paderborn University, the Georgia Institute of Technology, the University of Giessen, Washington