Sophie Langer

Email: s.langer(at)utwente.nl

University of Twente

Drienerlolaan 5

7522 NB Enschede

The Netherlands

I am an Assistant professor in the Statistics group at University of Twente. 

In 2020, I completed my PhD at the Technical University of Darmstadt under the supervision of Michael Kohler. 

After graduating, I worked as a postdoctoral researcher with Johannes Schmidt-Hieber at University of Twente and with Lorenzo Rosasco at Universita' di Genova, before joining University of Twente as an Assistant professor in October 2022.

My research mainly focus on the mathematical understanding of deep learning, combining statistical and computational-related aspects, with particular focus on deep learning in image classification. Recently, I got also interested in analysing different optimisation schemes and regularisation methods used in deep learning's training like dropout and SAM.

News and Upcoming Events

Since May 2024, I am a member of the Young Academy at University of Twente

Since March 2024, I am an associate editor for the Journal of Statistical Planning and Inference

Invited session organization at 6th International Symposium on Nonparametric Statistics in Braga, Portugal (2024/06)

Invited short course about Statistical theory of deep learning, Universidade de Aveiro, Portugal (2024/07)

Keynote speaker at the" Workshop for Junior Female Researchers in Probability" at TU Berlin (2024/07)

Invited talk at the workshop "Machine Learning in Infinite Dimensions" at University of Bath, UK (2024/08)

Scientific event on probabilistic and statistical analysis of random networks, stochastic processes and deep neural networks, Nice, France (2024/09)

Invited talk at the workshop Statistical Foundation of Deep Neural Network Models at  Brin Mathematics center, University of Maryland, USA (2024/11)

Oberwolfach Workshop "Frontiers of Statistics and Machine Learning" (03/2025)

Preprints

[1] Accelerated Mirror Descent for Non-Euclidean Star-convex Functions (with C. Lezane and W. Koolen). Preprint (2024).

[2] Dropout Regularization Versus l2-Penalization in the Linear Model (with G.Clara and J. Schmidt-Hieber). Preprint (2023).

[3] Statistical analysis of an image classification problem (with J. Schmidt-Hieber). Preprint (2022).

Publications

[1] Convergence rates for shallow neural networks learned by gradient descent (with A. Braun, M. Kohler and H. Walk). Bernoulli, 30(1): 475-502 (2024).

[2] Statistical theory for image classification using deep neural networks with cross entropy loss. Journal of Statistical Planning and Inference, to appear (2024).

[3] Learning green’s function efficiently using low-rank (with K. Wimalawarne and T. Suzuki). SynS ML @ ICML2023

[4] Estimation of a regression function on a manifold by fully connected deep neural networks (with M. Kohler and U. Reif). Journal of Statistical Planning and Inference, 222: 160-181 (2023).

[5] Estimation of a function of low local dimensionality by deep neural networks (with M. Kohler and A. Krzyzak). IEEE Transactions on Information Theory, 68(6): 4032-4042 (2022)

[6] Analysis of the rate of convergence of fully connected deep neural network regression estimates with smooth activation function. Journal of Multivariate Analysis, 182(C) (2021)

[7] Approximating smooth functions by neural networks with sigmoid activation function. Journal of Multivariate Analysis, 182(C) (2021)

[8] On the rate of convergence of fully connected deep very deep neural network estimates (with M. Kohler). Annals of Statistics, 49(4): 2231-2249 (2021)

[9] Discussion of "Nonparametric regression using deep neural networks with ReLU activation function" (with M. Kohler). Annals of Statistics, 48(4):1906-1910 (2020)

[10] Ein Beitrag zur statistischen Theorie des Deep Learnings. Verlag Dr. Hut (2020). 

[A] The Smoking Gun: Statistical theory improves neural network estimates (joint with Michael Kohler), Oberwolfach Report 2021

[B] The Role of Statistical Theory in Understanding Deep Learning, Oberwolfach Report 2023

Invited Talks

Oberwolfach Workshop "Applied Harmonic Analysis and Data Science" (2024/04)

Noon Seminar of C3S , Frankfurt, Germany (2024/04)

EURANDOM 25th anniversary , Eindhoven, Netherlands (2024/04)

SMSA, Delft, Netherlands (2024/03)

SIAM Conference on Uncertainity Quantification ,Trieste, Italy (2024/02)

Google Research Workshop on Deep Learning Down Under in Lorne, Australia (2024/01)

IMS-APRM , Melbourne, Australia (2024/01)

Oberwolfach Workshop "Nonlinear approximation of high-dimensional functions in scientific computing" (2023/10)

Latin American congress of probability and mathematical statistics, São Paulo (2023/07, link to program)

Keynote speaker at the International Conference of Maths4DL, London (2023/07)

Seminar++ - Machine Learning Theory,  Amsterdam (2023/05, link to program)

Workshop on Functional Inference and Machine Intelligence, Tokyo (2023/03, link to program)

Meeting in Mathematical Statistics, CIRM, Luminy (2022/12, link to program)

One World Seminar Series on the Mathematics of Machine Learning (2022/10, link to program, video of my talk)

Non-linear and High Dimensional Inference, IHP, Paris (2022/10, link to program and video)

IMS Annual Meeting, London (2022/06, link to program)

Curve and Surfaces, Arcachon (2022/06, link to program)

ESI Workshop on Computational Uncertainty Quantification: Statistical estimation and deep learning in UQ for PDEs, Vienna (2022/05, link to program, video of my talk)

CMStatstics, London (2021/12)

Oberwolfach Workshop: "Mathematical Foundations of Machine Learning" (2021/03)

Selected Seminar Talks: Karlsruher Institut für Technologie (2024/02), Ruhr-Universität Bochum (2023/12), Universität Heidelberg (2023/07), Eindhoven University of Technology (2023/01), UCLouvain (2022/12), Universität Leipzig (2022/11), Universität Heidelberg (2022/08), Technische Universität Braunschweig (2022/06), Humboldt-Universität Berlin (2022/05), Friedrich-Alexander-Universität Erlangen-Nürnberg (2022/02)

Awards and Grants

Incentive grant (together with Moritz Hahn and Janusz Meylahn), University of Twente, 2023 

Ruth Moufang Price - Best female PhD-student in Mathematics, TU Darmstadt, 2020

Ruth Moufang Price - Best female Postdoc in Mathematics, TU Darmstadt, 2020

Short Courses and Invited Lectures

Invited short course at the Latin American congress of probability and mathematical statistics, Sao Pãulo (2023/07)

Invited Lecture on Statistical Learning Theory and Applications at the Centre of Brains, Minds and Machines (link to course description), MIT, Boston (2022/11)

Research school (part 1): High-dimensional approximation and Deep Learning (link to program), Nantes (2022/05)

Online short course on (deep) statistical learning, UCLouvain (2021/12)

Online short course on Statistical Theory of Deep Learning, International Statistical Institute (2022/12)

Teaching

Capita Selecta Statistics on Theory of Reinforcement Learning, University of Twente (summer semester 2024)

Statistics for Industrial Engineering and Management, University of Twente (summer semester 2024)

Mathematical Statistics II, University of Twente (fall semester 2023/2024)

Capita Selecta Statistics on the Theory of Deep Learning (link to course description),  University of Twente (summer semester 2023)

Lecture on Statistical Foundations of Deep Learning, TU Darmstadt (summer semester 2021)

Conference and Seminar Organizations

Seminar series on the Mathematics of Data Science (with Marcello Carioni),  University of Twente  (since 2023/01-)

Session organization (with Tim van Erwen), Second workshop on AI and Mathematics, University of Twente (2023/06)

Machine Learning Session (with Michael Kohler), German Probability and Statistic Days 2023, Essen (2023/03)

Workshop on the statistical analysis of deep learning (with Stefan Richter and Johannes Schmidt-Hieber), Sommerakadamie der Studienstiftung, Hattingen (2022/08)

Minisymposium on "The Role of Learning Theory in Machine Learning" (with Nicole Mücke), EWM General Meeting, Helsinki (2022/08)

Miscellaneous

The invariable mark of wisdom is to see the miraculous in the common.

Ralph Waldo Emerson