Skip to main content

Making machine learning models less data-hungry: Vincent Fortuin acquires Branco Weiss Fellowship

Awards & Grants, Health AI,

Helmholtz Munich researcher Dr Vincent Fortuin is one of the young investigator group leaders at Helmholtz AI. His work focuses on the interface between Bayesian inference and deep learning. Bayesian inference is a statistical method to learn from data by combining explicit prior knowledge with the data. The goal of the research group is to improve robustness, data efficiency, and uncertainty estimation in modern machine learning approaches. Vincent Fortuin has now acquired the competitive Branco Weiss Fellowship for his research on Bayesian statistics, representation learning, and meta-learning.

Bayesian statistics help integrate prior knowledge into deep learning processes. Usually, artificial intelligence (AI) learns blindly from a huge number of samples, but in many applications, those big datasets are not available. However, for many of these subjects, there is a lot of theoretical knowledge available. This information could be integrated into the AI before training, allowing a better interpretation of the available data even if it’s scarce. Bayesian statistics can help identify how to optimally use prior knowledge in any learning problem so that AI models can be trained efficiently using way fewer data.

The goal of Vincent Fortuin’s research group is to study the role of prior knowledge in Bayesian deep learning, working towards more data-efficient learning approaches targeted at critical applications, where model robustness and calibrated uncertainty estimates are crucial. Some areas where Bayesian machine learning could be used include intensive care medicine, single-cell multi-omics, and drug design.

As a Branco Weiss Fellow, Vincent Fortuin will be supported in his work in Bayesian deep learning, aiming to unlock new applications for areas where only small datasets are available. The competitive grant searches for excellence and has created a community of exceptional junior researchers from a large spectrum of fields and countries.