Technische Universität München, Department of Mathematics
Lecture
First Order Primal-Dual Optimization Methods
Winter Semester 2013/14
Contents of the Lecture
First-order methods have recently emerged into a highly active research field in optimization. They usually combine primal-dual
decomposition approaches with relatively simple iteration schemes. This makes them very well suited for designing powerful,
structure-exploiting algorithms for challenging large-scale applications. This course gives an advanced introduction into the design
and theory of first-order primal-dual optimization methods. The module description can be found
here.
News
November 8th, 2013: |
Group 2 will take place on November 19th, instead of November 12th, due to the Studentische Vollversammlung (SVV). |
November 5th, 2013: |
The hint in exercise 2.2 d) has been updated. |
November 4th, 2013: |
The second exercise sheet is online. |
November 4th, 2013: |
The lecture on Wednesday, Nov 6th, 10:15-11:45, had to be canceled. We will look for an alternative date. The next lecture is on Wednesday, Nov 13th, 10:15-11:45. |
October 30th, 2013: |
Lecture notes updated. |
October 25th, 2013: |
The Doodle is now closed. It shows that unfortunately there is no better time for group 2 than Tuesday 10:15-11:45, hence the date remains unchanged. |
October 24th, 2013: |
Please participate in the Doodle until friday, october 25th, 12:00, or otherwise your desired times cannot be considered. |
October 23rd, 2013: |
A Doodle to determine the best suitable time for group 2 has been set up. More info under Dates. |
October 21st, 2013: |
The first exercise sheet is now online. It will be discussed in the exercises on Oct 24th and in group 2. |
October 18th, 2013: |
It is now possible to sign up for exercise groups 1 and 2 in TUMonline. |
October 9th, 2013: |
The first lecture takes place on Wednesday, October 16th. The exercises will start in the week after that. |
Dates
Exam
Place and date of the exam are not yet fixed and will be annonuced here as soon as they are available.
More information will follow.
Lecture Notes
- Current version of Lecture Notes (will be dynamically updated as the course evolves).
Exercises
- Responsible for the exercises is Dennis Klöckner. In case of questions or suggestions, please feel encouraged to contact Dennis.Kloeckner
ma.tum.de.
- The exercises take place biweekly.
- Handwritten solution sketches of the exercises will be provided here.
- Exercise Sheets:
References
- Stephen Boyd, Neal Parikh, Eric Chu, Borja Peleato, Jonathan Eckstein: Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers, Foundations and Trends in Machine Learning Vol. 3, No. 1 (2010), 1–122.
- Antonin Chambolle, Thomas Pock: A First-Order Primal-Dual Algorithm for Convex Problems with Applications to Imaging, Journal of Mathematical Imaging and Vision, Vol. 40, No. 1 (2011), 120-145.
- Thomas Pock, Daniel Cremers, Horst Bischof, Antonin Chambolle: An Algorithm for Minimizing the Piecewise Smooth Mumford-Shah Functional. In IEEE International Conference on Computer Vision (ICCV), 2009.