Prof. Dr.
Michael
Vogt
Institute of Statistics
Helmholtzstraße 20
89081
Ulm
Baden-Württemberg
Germany
Dozent: Michael Vogt
Übungsleiter: Manuel Rosenbaum
Allgemeine Informationen:
Vorlesung: | 2 h (weitere Informationen siehe Moodle/LSF) |
Übung: | 1 h (weitere Informationen siehe Moodle/LSF) |
Inhalt: | 1. Classification and Regression Problems - Statistical decision theory - Binary classification - Logistic regression 2. Technical Tools - Exponential inequalities - Concentration inequalities - Subgaussian random variables 3. Uniform Convergence and Generalization - Classification with 0-1-loss - Convex relaxation for classification - Regression 4. Neural Networks - Definition of deep neural networks - Statistical model - Generalization bounds based on Rademacher complexity - Approximation theory - Convergence rates |