Resource-Efficient Bayesian Machine Learning

• Applicant: Professor Dr. Robert Bamler
• Subject Area: Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
• Term: since 2021
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 448588364

Rapid recent progress in machine learning has made the field a pillar of Artificial Intelligence research. However, many of these achievements have been fueled by labor-intensive generation of massive data sets and enormous investments into computational resources. This proposal aims to develop machine learning methods that are resource efficient in terms of training data, storage, bandwidth, and computing power. The developed methods will broaden the reach of machine learning to new and emerging fields. An emphasis will be on applications of machine learning to natural sciences (where training data is often expensive), to the young and dynamic field of neural data compression (which increases efficiency in bandwidth and storage), and to a vision towards a radically new form of machine learning on authority-free decentralized computing platforms (where computational power and communication are very limited). These three application domains will provide concrete test beds for foundational research on general purpose methods from the statistical machine learning domain, in particular probabilistic models and approximate Bayesian inference, which provide an overarching theme of the proposal.

• DFG Programme: Independent Junior Research Groups
• International Connection: USA
• Cooperation Partner: Professor Dr. Stephan Mandt