[w20/21] Seminar: Manifold Learning - Modern Approaches for Dimensionality Reduction

One of the most challenging aspects of dealing with the ongoing "big data" explosion is the development of methods that can identify suitable low-dimensional representations of very high dimensional data sets. The unifying assumption of all such approaches is that high-dimensional data are concentrated in a lower-dimensional subspace (a "manifold", more generally) embedded in the original data space.

In this seminar, we will introduce the mathematical basics of manifolds and embeddings and will discuss both foundational papers on popular dimensionality reduction methods and papers on current research problems in this setting.

General Information

• Kickoff Meeting: 30.10.2020, 12:00 - 14:00
  
• Due to the current situation the seminar will be held online.
• Enrolment key: manil2021