e-Learning Support
Seminar: 03-M-AC-35 Large Scale Convex Optimization - Details

Seminar: 03-M-AC-35 Large Scale Convex Optimization - Details

Nachrichtenbox schließen
Sie sind nicht in Stud.IP angemeldet.

Allgemeine Informationen

Veranstaltungsname Seminar: 03-M-AC-35 Large Scale Convex Optimization
Untertitel
Veranstaltungsnummer 03-M-AC-35
Semester SoSe 2025
Aktuelle Anzahl der Teilnehmenden 12
Heimat-Einrichtung Mathematik
Veranstaltungstyp Seminar in der Kategorie Lehre
Erster Termin Mittwoch, 09.04.2025 12:00 - 14:00, Ort: MZH 1100
Art/Form
Englischsprachige Veranstaltung Ja
ECTS-Punkte 4,5/6

Räume und Zeiten

MZH 1100
Mittwoch: 12:00 - 14:00, wöchentlich (14x)
MZH 3150
Dienstag, 24.06.2025 10:00 - 12:00
MZH 4140
Donnerstag, 26.06.2025 10:00 - 12:00

Modulzuordnungen

Kommentar/Beschreibung

This seminar will treat methods of convex optimization that are suitable for large problems with millions of variables (the size is basically restricted that the computer memory can hold a small finite number of vectors). The guiding principle of these methods is to exploit suitable additive \emph{splitting} of the objective function and then use simple building blocks for the splitted parts to assemble an algorithm. Recently there has been much progress in this areas and in this seminar we will explore the following directions, for example: \begin{itemize} \item Stochastic optimization (stochastic gradient descent, stochastic proximal gradient method), [Chapter 7, 1] \item Accelerated methods (Accelerated gradient descent, accelerated proximal point methods), [Chapter 12, 1] \item Degenerate preconditioned proximal point methods [2] \item Kaczmarz (row-action) methods with mismatched adjoint [3] \end{itemize}