Universität Regensburg   IMPRESSUM   DATENSCHUTZ
Fakultät für Mathematik Universität Regensburg
Nonsmooth Optimization
Semester
SoSe 2022

Lecturer
Luise Blank

Type of course (Veranstaltungsart)
Seminar

German title
Nichtglatte Optimierung

Contents
The seminar covers optimization problems which do not satisfy the typical differentiability requirements. Starting from convex functions, the notion of subdifferentials is introduced and with it optimality conditions for unrestricted problems are derived. Allowing infinity as a function value, restrictions can be included with the help of the indicator function. Based on these regularization procedures and subgradient methods for the numerical solution of non-smooth optimization problems are discussed. Also a generalized (non-smooth) Newton method is introduced.

Literature
  • Geiger, Kanzow: Theorie und Numerik restringierter Optimierungsaufgaben,
  • Kiwiel: Methods of descent for nondifferentiable optimization,
  • and some articles


Recommended previous knowledge
Numeric I, Optimization I is helpful but not required

Time/Date
Tue 10-12

Location
M101

Registration
  • Organisational meeting/distribution of topics: Tuesday 1.2.22, 11:45 in M101
  • Registration for course work/examination/ECTS: FlexNow
Course work (Studienleistungen)
  • Presentation: Giving a seminar talk of roughly 90 minutes
Examination (Prüfungsleistungen)
  • Detailed written report of the seminar talk
Modules
BSem, MV, MSem, LA-GySem

ECTS
BSem und MSem: 4,5 LP bei Studienbeginn ab WS 15/16, 6 LP bei Studienbeginn vor WS 15/16. LA-GySem:
6 LP. MV und Nebenfach: 4,5 LP bei Studienbeginn ab WS 15/16, 6 LP bei Studienbeginn vor WS 15/16
+++ weitere Details: siehe Modulkatalog +++