Project
EF1-8
Incorporating Locality into Fast(er) Learning
Project Heads
Nicole Mücke
Project Members
Abhishake Rastogi
Project Duration
01.10.2020 − 30.09.2022
Located at
TU Berlin
Description
Classical stochastic approximation methods such as SGD in reproducing kernel Hilbert spaces are not able to exploit regions of different regularity of the target function. This slows down local convergence dramatically. To overcome this drawback, we propose to analyze localized SGD approaches.
Project Webpages
Selected Publications
[1] Nicole Mücke, Enrico Reiss, Stochastic Gradient Descent in Hilbert Scales: Smoothness, Preconditioning and Earlier Stopping, arXiv:2006.10840
[2] Nicole Mücke, Gergely Neu, Lorenzo Rosasco, Beating SGD Saturation with Tail-Averaging and Minibatching, 33rd Conference on Neural Information Processing Systems (NeurIPS 2019), Vancouver, Canada,
http://papers.nips.cc/paper/9422-beating-sgd-saturation-with-tail-averaging-and-minibatching
[3] Nicole Mücke, Reducing training time by efficient localized kernel regression, Proceedings of Machine Learning Research, PMLR 89:2603-2610, 2019.
http://proceedings.mlr.press/v89/muecke19a.html
[4] Nicole Mücke, Stochastic Gradient Descent Meets Distribution Regression, Proceedings of the 24th International Conference on Artificial Intelligence and Statistics (AISTATS) 2021, San Diego, California, USA. PMLR: Volume 130.
http://arxiv.org/abs/2010.12842
[5] Ernesto De Vito, Nicole Mücke, Lorenzo Rosasco, Reproducing kernel Hilbert spaces on manifolds: Sobolev and Diffusion spaces, Analysis and Applications 2020.
https://www.worldscientific.com/doi/abs/10.1142/S0219530520400114
[6] Nicole Mücke, Enrico Reiss, Jonas Rungenhagen, Markus Klein, Data splitting improves statistical performance in overparametrized regimes, arXiv:2110.10956
Selected Pictures
Please insert any kind of pictures (photos, diagramms, simulations, graphics) related to the project in the above right field (Image with Text), by choosing the green plus image on top of the text editor. (You will be directed to the media library where you can add new files.)
(We need pictures for a lot of purposes in different contexts, like posters, scientific reports, flyers, website,…
Please upload pictures that might be just nice to look at, illustrate, explain or summarize your work.)
As Title in the above form please add a copyright.
And please give a short description of the picture and the context in the above textbox.
Don’t forget to press the “Save changes” button at the bottom of the box.
If you want to add more pictures, please use the “clone”-button at the right top of the above grey box.