Name: Dr. Frank Schneider
Title: Postdoctoral Researcher
Affiliation: University of Tübingen, Germany
Bio:
I’m a
postdoctoral researcher at the University of Tübingen in the
Methods of Machine Learning group, led by
Philipp Hennig. I’m also a chair of the
Algorithms working group at MLCommons.
My research focuses on
efficient and user-friendly training methods for machine learning. I’m particularly interested in eliminating tedious hyperparameters (e.g., learning rates, schedules) to automate neural network training and make deep learning more accessible. A key aspect of my work is designing rigorous and meaningful benchmarks for training methods, such as
AlgoPerf.
Previously, I earned my PhD in Computer Science from the University of Tübingen, supervised by Philipp Hennig, as part of the
IMPRS-IS (International Max Planck Research School for Intelligent Systems). Before that, I studied
Simulation Technology (B.Sc., M.Sc.) at the University of Stuttgart and
Industrial and Applied Mathematics (M.Sc.) at TU/e Eindhoven. My master’s thesis, supervised by
Maxim Pisarenco and
Michiel Hochstenbach, explored novel preconditioners for structured Toeplitz matrices. This thesis was conducted at ASML (Eindhoven), a company specializing in lithography systems for the semiconductor industry.
Twitter
LinkedIn