Statistical and machine learning approaches that deal very successfully with uncertain data in complex disciplines such as neuroscience, medicine, and artificial intelligence (AI) – known as Bayesian methods – may struggle under additional practical constraints like costly model evaluations and the need to preserve the privacy of data subjects. The WiseBayes consortium project will exploit a unique combination of expertise at the University of Helsinki to produce a new generation of machine learning methods for Bayesian inference which are simultaneously able to take into account resource costs (e.g., time, energy, compute), privacy concerns, and accurate uncertainty quantification. Thanks to this resource-wise and trustworthy approach, the WiseBayes project will promote sustainable machine learning and advance open science with algorithms for data analysis and data sharing that are more widely applicable than before while being respectful of individual data privacy.

2023

2027