This lecture introduces the concept of Rademacher complexity, which is frequently used in machine learning as a means to measure the richness or capacity of function classes in a given learning problem. By proving bounds on Rademacher complexities of affine function classes and by outlining the relevance of these bounds for classes of neural networks, we show how Rademacher complexities are employed and why they are useful.
Introduction to Rademacher complexities
19.01.2021 14:00 - 14:20
Organiser:
Fakultät für Mathematik
Location:
Online via Zoom