Independent Research Fund Denmark is granting DKK 53.4 million to research focusing on the responsible development and use of artificial intelligence, aiming to strengthen Denmark's position in the international competition for AI development.
Jes Frellsen receives the grant for his research project 'Reclaiming uncertainty: Robust Bayesian deep learning by handling overparameterization'. With the project, he aims to make AI more reliable by combining so-called deep neural networks, models that can identify patterns in complex data, with a hundred-year-old mathematical approach that makes it possible to calculate how uncertain a model is in its predictions. However, neural networks are so complex and full of symmetries that this classical approach cannot be applied as it is. Jes Frellsen therefore aims to develop new mathematical and algorithmic tools that consider the networks' structure and symmetries, thereby achieving more reliable uncertainty estimates.
