Lecture: Convex Optimization and Online Learning (MM25)
Language: English or German, as the audience requests.
Content: This lecture covers basic concepts of convex analysis and programming (classes of convex sets and functions, duality, nonexpansive mappings, splitting of convex programs, etc) with a focus on the analysis of iterative algorithms for solving large-scale convex optimization problems. Particular attention is paid to techniques for online convex optimization and the prediction of individual sequences in unknown environments, which play a key role in machine learning applications.
The content of the lecture is targeted at students of mathematics and scientific computing with a long-term interest in machine learning, to prepare them for more advanced topics closer to research.
Prerequisites: All proofs are elementary and only require knowledge from the mandatory undergraduate courses on analysis and linear algebra.
Registration: If you wish to attend the lecture and the exercises, please sign up using MÜSLI.
You have to be logged in to access the files listed below.
Lecture Notes: All statements (second part) will be proven in the lecture.
Convex Analysis and Programming
- shortcut: KKT Conditions
- Projection (update: 22.11.18), Proximal Mappings (update: 04.12.18)
Online Convex Optimisation and Learning
- Sheet 1 (TBD 31.10)
- Sheet 2 (TBD 07.11)
- Sheet 3 (TBD 14.11, exercise 5: 21.11) / data_ex5.zip, code_ex3.5.zip
- Sheet 4 (TBD 21.11)
- Sheet 5 (TBD 28.11) / code_ex5.3.zip
- Sheet 6 (TBD 05.12)
- Sheet 7 (TBD 12.12) / data_ex3.zip , code_ex7.3.zip
- Sheet 8 (TBD 19.12, exercise 4: 09.01) / data_ex4.zip, code_ex8.4.zip
- Sheet 9 (TBD 09.01)
- Sheet 10 (TBD 16.01) / code_ex10.4.zip
- Sheet 11 (TBD 23.01) / data_ex11.2.zip
- Sheet 12 (TBD 30.01)