Prof. Frank Hutter
|
|
Gruppenleiter Georges-Köhler-Allee 7479110 Freiburg im Breisgau Gebäude 74, Raum 00-017 |
Automatic machine learning (AutoML)
We aim to democratize machine learning by means of open-source software that allows even non-experts to use effective machine learning solutions for their data. We are multiple-time world champions in the AutoML challenge series.
Deep learning
We particularly focus on efficient training algorithms, efficient neural architecture search and hyperparameter optimization, and Bayesian deep learning.
Automated algorithm design
We develop automated methods for algorithm configuration, algorithm selection, and algorithm analysis.Typical application domains are NP-hard problems, such as SAT solving, AI planning, or TSP solving.
Bayesian Optimization
This entails sequential experimental design under uncertainty for sample-efficient global zeroth order optimization.