Publica
Hier finden Sie wissenschaftliche Publikationen aus den FraunhoferInstituten. Towards largescale gaussian process models for efficient bayesian machine learning
 Hammoudi, Slimane (Ed.) ; Institute for Systems and Technologies of Information, Control and Communication INSTICC, Setubal: 9th International Conference on Data Science, Technology and Applications 2020. Proceedings : 7  9 July, 2020, webbased event Setúbal: INSTICC, 2020 ISBN: 9789897584404 S.275282 
 International Conference on Data Science, Technology and Applications (DATA) <9, 2020, Online> 

 Englisch 
 Konferenzbeitrag 
 Fraunhofer FIT () 
Abstract
Gaussian Process Models (GPMs) are applicable for a large variety of different data analysis tasks, such as time series interpolation, regression, and classification. Frequently, these models of bayesian machine learning instantiate a Gaussian Process by a zeromean function and the wellknown Gaussian kernel. While these default instantiations yield acceptable analytical quality for many use cases, GPM retrieval algorithms allow to automatically search for an applicationspecific model suitable for a particular dataset. Stateoftheart GPM retrieval algorithms have only been applied for small datasets, as their cubic runtime complexity impedes analyzing datasets beyond a few thousand data records. Even though global approximations of Gaussian Processes extend the applicability of those models to mediumsized datasets, sets of millions of data records are still far beyond their reach. Therefore, we develop a new largescale GPM structure, which incorporates a divide& conquerbased paradigm and thus enables efficient GPM retrieval for largescale data. We outline challenges concerning this newly developed GPM structure regarding its algorithmic retrieval, its integration with given data platforms and technologies, as well as crossmodel comparability and interpretability.