Estimating causal effects using NADEs

Sergio Garrido is a Ph.D. student at the Transport department at the Technical University of Denmark, working on solving transport problems using machine learning methods. He is mainly interested in probabilistic models, causality, and neural networks. Currently, he is a Ph.D. research intern at the causality lab at Amazon, Tübingen.

During the presentation, he will introduce us to causality from Pearl's perspective and a class of deep generative models called neural autoregressive density estimators (NADEs). After introducing both concepts, he will explain how they can be naturally combined and show some proof-of-concept results against basic benchmarks of the approach.