IAS NTU Discovery Science Seminar: Applied Causality in Computer Vision by Prof Hanwang Zhang

You must have heard of the topos: correlation is not causality; or seeing is illusion. But, except for seeing and then measuring the correlation between data, what else can we do to seek causality? In this talk, Prof Zhang showed that there are indeed many more we can do, but in a different way of "seeing’’ and "measuring”. They are:
1) How to think causally? It is a joyful and non-technical introduction of causality, to help you get to know how to think in a causal way.
2) How to capture causality? He introduced some recent works on causal interventions, that is, "seeing’’ differently can make it causal.
3) How to imagine using causality? For years, we care much about generalization. In fact, it is all about imagining the unseen scenario. He introduced recent works on counterfactuals.
4) How to go beyond? What is the cause of the causality? Perhaps, answering this is the true future of machine learning.

Видео IAS NTU Discovery Science Seminar: Applied Causality in Computer Vision by Prof Hanwang Zhang канала NTU- Institute of Advanced Studies