PhD scholarship in Geometric Bayesian Deep Learning – DTU Compute

Do you want to figure out why Bayesian deep learning doesn’t work? And afterwards fix it? At DTU Compute we are working towards building highly scalable Bayesian approximations that actually work for deep learning. The driving approach is to first figure out the root cause of why the usual tricks do not work, and then secondly fix the issues. The research environment favors creative ‘slow thinking’ style research, and we emphasize having fun along the way. Our work is consistently published at the top venues of machine learning research.