Universität Wien

053612 VU Optimisation Methods for Data Science (2020W)

Prüfungsimmanente Lehrveranstaltung

An/Abmeldung

Hinweis: Ihr Anmeldezeitpunkt innerhalb der Frist hat keine Auswirkungen auf die Platzvergabe (kein "first come, first served").

Details

max. 25 Teilnehmer*innen
Sprache: Englisch

Lehrende

Termine (iCal) - nächster Termin ist mit N markiert

  • Donnerstag 08.10. 11:30 - 14:45 Digital
  • Donnerstag 15.10. 11:30 - 13:00 Digital
  • Donnerstag 22.10. 11:30 - 14:45 Digital
  • Donnerstag 29.10. 11:30 - 13:00 Digital
  • Donnerstag 05.11. 11:30 - 14:45 Digital
  • Donnerstag 12.11. 11:30 - 13:00 Digital
  • Donnerstag 19.11. 11:30 - 14:45 Digital
  • Donnerstag 26.11. 11:30 - 13:00 Digital
  • Donnerstag 03.12. 11:30 - 14:45 Digital
  • Donnerstag 10.12. 11:30 - 13:00 Digital
  • Donnerstag 17.12. 11:30 - 14:45 Digital
  • Donnerstag 07.01. 11:30 - 13:00 Digital
  • Donnerstag 14.01. 11:30 - 14:45 Digital
  • Donnerstag 21.01. 11:30 - 13:00 Digital
  • Donnerstag 28.01. 11:30 - 14:45 Digital

Information

Ziele, Inhalte und Methode der Lehrveranstaltung

Full digital synchronous mode (e-meet the professors live every week; recording is not guaranteed!)

Platform is moodle:

https://moodle.univie.ac.at/course/view.php?id=169428

--------------------
Contents:

1. Geometric foundations of duality

1.1 Convexity and minimal distance projection
1.2 Properties of the minimal distance projection
1.3 Separation of convex sets
1.4 Supporting hyperplane and Farkas' Lemma

2. The concept of duality in optimization

2.1 Lagrange duality for constrained optimization problems
2.2 Duality gap, quality guarantee, and complementary slack
2.3 Minimax, saddle points, and optimality conditions
2.4 Convex problems: Slater condition, Wolfe dual

3. Practical aspects of duality in optimization

3.1 Linear and quadratic optimization
3.2 Ascent directions for the dual function
3.3 Dual (steepest) ascent method
3.4 (Dual) cutting planes
3.5 Duality for discrete problems; branch-and-bound

Art der Leistungskontrolle und erlaubte Hilfsmittel

(1) virtual-oral presentations of exercises (from the lecture notes, to be prepared in advance; format: a single .pdf with your name, max.size 5MB) which will be awarded by up to 30 points.

(2) a take-home exam (scheduled by majoritry vote to 14 January 2021). Net working time will be set tight, so we suggest to prepare well (from experience, you will lack time to find the answer during exam without having thought of the topic before). Details will be communicated in due course.
Exam will be awarded by up to 20 points.

(3) Active virtual cooperation during class will be awarded by up to 20 points, depending on the intensity and relevance of your communication (e.g., questions regarding administration won't be relevant for grading)

(4) To pass the exam/course successfully, you need 36 points.

Grades:

0-35: nicht genuegend/fail (5)
36-43: genuegend/pass (4)
44-53: befriedigend/satisfactory (3)
54-63: gut/good (2)
64-70: sehr gut/excellent (1)

Mindestanforderungen und Beurteilungsmaßstab

see above

Prüfungsstoff

all material covered by lecture notes (see moodle)

Literatur

Lecture notes

Bazaraa, M.S., Sherali, H.D., Shetty, C.M.: Nonlinear Programming: Theory and Algorithms, Wiley

Zuordnung im Vorlesungsverzeichnis

Modul: OMD

Letzte Änderung: Fr 12.05.2023 00:13