Project
AA5-1 (was EF1-14)
Sparsity and Sample-Size Efficiency in Structured Learning
Project Heads
Sebastian Pokutta
Project Members
David Martínez-Rubio
Project Duration
01.01.2022 − 31.12.2023
Located at
ZIB
Description
In this project we have made important advances in the development of optimization algorithms designed for high-dimensional tasks, in several directions. We have developed a fast accelerated algorithm for the PageRank problem for finding local sparse clusters in a time that depends on the fi nal cluster despite of not knowing it in advance. We have made several contributions in Riemannian optimization, using the geometric structure of the problems and have obtained fast fi rst-order optimization methods for machine learning tasks, such as a generalization of Nesterov’s accelerated gradient descent to the manifold setting and accelerated algorithms for min-max Riemannian problems. We have also developed fast optimization algorithms for the computation of a fair solution in a packing problem and in its dual, applicable to fair resource allocation and also to general linear programming.
External Website
More detailed information about this project can be found on its external homepage.