Institute Homepage
Institute Homepage DE Sign In

Back

Intelligent Control Systems Conference Paper 2018

Learning Gaussian Processes by Minimizing PAC-Bayesian Generalization Bounds

Thumb ticker sm doerr andreas 09 croped
Intelligent Control Systems
Andreas Doerr
Pac gp img2

Gaussian Processes (GPs) are a generic modelling tool for supervised learning. While they have been successfully applied on large datasets, their use in safety critical applications is hindered by the lack of good performance guarantees. To this end, we propose a method to learn GPs and their sparse approximations by directly optimizing a PAC-Bayesian bound on their generalization performance, instead of maximizing the marginal likelihood. Besides its theoretical appeal, we find in our evaluation that our learning method is robust and yields significantly better generalization guarantees than other common GP approaches on several regression benchmark datasets.

Author(s): Reeb, David and Doerr, Andreas and Gerwinn, Sebastian and Rakitsch, Barbara
Book Title: Proceedings Neural Information Processing Systems
Year: 2018
Month: December
Bibtex Type: Conference Paper (inproceedings)
Event Name: Neural Information Processing Systems (NIPS 2017)
Event Place: Montreal, Canada
Electronic Archiving: grant_archive

BibTex 

@inproceedings{learning_pac_gp,
  title = {Learning Gaussian Processes by Minimizing PAC-Bayesian Generalization Bounds},
  booktitle = {Proceedings Neural Information Processing Systems},
  abstract = {Gaussian Processes (GPs) are a generic modelling tool for supervised learning. While they have been successfully applied on large datasets, their use in safety critical applications is hindered by the lack of good performance guarantees. To this end, we propose a method to learn GPs and their sparse approximations by directly optimizing a PAC-Bayesian bound on their generalization performance, instead of maximizing the marginal likelihood. Besides its theoretical appeal, we find in our evaluation that our learning method is robust and yields significantly better generalization guarantees than other common GP approaches on several regression benchmark datasets.},
  month = dec,
  year = {2018},
  slug = {learning_pac_gp},
  author = {Reeb, David and Doerr, Andreas and Gerwinn, Sebastian and Rakitsch, Barbara},
  month_numeric = {12}
}