Noon Talk with Sophie Langer
Abstract
In recent years, there has been a surge of interest across different research areas to improve the theoretical understanding of deep learning. A very promising approach is the statistical one, which interprets deep learning as a nonlinear or nonparametric generalization of existing statistical models. For instance, a simple fully connected neural network is equivalent to a recursive generalized linear model with a hierarchical structure. Given this connection, many papers in recent years derived convergence rates of neural networks in a nonparametric regression or classification setting. Nevertheless, phenomena like overparameterization seem to contradict the statistical principle of bias-variance trade-off. Therefore, deep learning cannot only be explained by existing techniques of mathematical statistics but also requires a radical overthinking. In this talk we will explore both, the importance of statistics for the understanding of deep learning, as well as its limitations, i.e., the necessity to connect with other research areas.
Bio
Sophie Langer is an assistant professor in the Statistics group at University of Twente and specializes on the theory of deep learning. Her research focuses on the statistical and computational aspects of neural networks, particularly on applications in image classification.
After receiving her PhD in Statistics from Technical University of Darmstadt under the supervision of Michael Kohler in 2020, Sophie worked as a PostDoctoral Researcher with Johannes Schmidt-Hieber (University of Twente) and Lorenzo Rosasco (University of Genoa). In October 2022, she joined the University of Twente as an assistant professor.
Sophie has also been a guest researcher in Tomaso Poggio’s group at MIT and Taiji Suzuki’s group at University of Tokyo.
She has given several invited lectures and short course on statistical aspects of deep learning, including at MIT, Boston and XVI Clapem, São Paulo. She currently serves as a Guest Editor for a special issue on “Deep Learning: Statistical Perspectives” in the Journal of Statistical Planning and Inference and has recently taken on the role of an Associate Editor since March 2024.
Zoom Registration
The talk will take place on Friday, April 19th, 12:00 CEST.
Please register with office@c3s.uni-frankfurt.de to receive the Zoom login.