$\vec{w}h\alpha\mathfrak{t}\;\; i\mathbb{S}\ldots$

What are...generative probability flows for sampling?


Who?
Viktor Stein (TU Berlin)
When?
2025/11/07, 13:00
Where?
TU Berlin (Math building, room MA 041, ground floor).
About what?

In this talk, I will give an elementary introduction to optimal transport and a metric space of probability measures called the Wasserstein space. After defining gradient flows on metric spaces, we will look at a certain PDE - the continuity equation - from an Eulerian and Lagrangian perspective. I will explain how these PDEs are used in machine learning, employed in generative models for sampling from a probability measure defined on a high-dimensional space.