Swedish web site
Optimal Transport Theory for Control and Machine Learning2026HT, 6.0 credits
|
|
Course plan
No of lectures
About 4 lectures
About 4 sessions in reading group format (depending on the number of students)
Recommended for
PhD students who have a background in machine learning, automatic control, applied mathematics, or related fields. The course is especially recommended for students who would like to explore using optimal transport theory in their research.
The course was last given
--
Prerequisites
- Good background in real analysis
- Some familiarity with measure theory and optimization is recommended
Organization
One 2-day block of lectures
One set of homework exercises
A reading group that will be organized either as a block or as a regular online
event
An optional project (for additional 3 credits)
Content
The course provides an introduction to optimal transport theory, covering the
following topics:
- Monge & Kantorovich formulation
- Duality theory
- Static and dynamic version (Brenier-Benamou formulation)
- Schrödinger bridges
- Gradient flows
Building on this introduction, students explore their own interests by reading
a paper of their choice, for which they lead a reading group discussion and
write a reflective report.
Optionally, students can do an additional self-defined project (+3 credits).
Literature
- Gabriel Peyré & Marco Cuturi, Computational Optimal Transport, 2019
- Lecture notes / slides
Lectures (Preliminary)
1. Monge & Kantorovich formulation
2. Duality theory
3. Static and dynamic formulation (Brenier Benamou formulation)
4. Schrödinger bridges & gradient flows
Examination
- One set of homeworks
- Leading a discussion in the reading group and writing a reflective
report
- Optionally, an additional project of the students’ choice
Examiner
Isabel Haasler
Credits
6 + 3 (optional project)
Comments
Page responsible: Anne Moe
