DSML Reading Club #7: causal representation learning

Causal Representation Learning from Multiple Distributions: A General Setting. In many problems, measured variables are functions of latent causal variables such as underlying concepts or objects. Recovering these latent variables and their causal relationships helps make predictions in changing environments and apply appropriate changes to a system. This is known as causal representation learning. The paper studies a general, fully nonparametric setting with multiple distributions and does not assume that distribution changes come from hard interventions.

At the meeting, we will discuss:

• What causal representation learning is
• How it can be applied, using EEG as an example
• Parametric and nonparametric settings and VAEs
• Saturday, September 13, 11:00 Kazakhstan time
• Add to calendar
• meet.google.com/aeo-oivb-zdp
• Speaker: Ayana Mussabayeva