The vanderschaar-lab presents “the map of causal deep learning (CDL)” as a way to formalise and categorise the many methods proposed under the causal deep learning name. By combining a structural scale as well as a parametric scale, the map bridges two worlds: causality and deep learning, respectively.
Our map is proposed as a living paper. This means that we welcome discussion and suggestions from the broader community. In the future, we will facilitate this through an online forum as well as an interactive version of our map.
We look forward to what the future of CDL holds!
In the meantime we refer to our research pillar on the topic where we introduced a “rung 1.5” to Pearl’s ladder of causation: