Postdoctoral Fellowship on Deep Probabilistic Models for Robust Representation Learning (m/f/d)

We expect this project to advance our understanding as well as the robustness and interpretability of probabilistic deep generative models. An improved analytical understanding of probabilistic models such as VAEs can help us to determine their limitations and push their boundaries. We are particularly interested in going beyond traditional statistical representations by moving towards representations that support notions of distribution change, intervention, and other forms of robustness. This will ultimately allow us to identify existing and novel mechanisms that lead to learning more versatile representations supporting artificial intelligence. It may also enable better interpretability and allow more meaningful interaction with humans.

Thumb ticker sm l1170153  0.5 credit
ELLIS Institute Tübingen & Max Planck Institute for Intelligent Systems