Learning Causal Representations From Unknown Interventions
summary
Claims identifiability for latent causal variables that participate in changing mechanisms and gives concrete tests and algorithms to recover them. Uses the GIN condition (a linear combination of Y independent of Z) to cluster measured variables and recover latent parents and causal order, validates on teacher burnout (28 variables) finding latent structure matching expert hypotheses, and leverages nonlinear ICA with changing component distributions to identify independent latent confounders; notes that latent factors whose correlations never change remain unrecoverable.
AI-written summary · about
chapters (15)
- 0:00Intro
- 1:03Causality and Modular Changes
- 4:25Outline
- 4:58Constraint-Based Causal Discovery
- 6:18Functional Causal Model-Based Cause Discovery
- 7:19In the Presence of Latent Confounde
- 8:53GIN for Estimating Linear, Nor Gaussian LV Moda
- 10:48Application to Teacher's Burnout Dat
- 11:19Estimating Latent Hierarchic Structure with GIL
- 14:31Learning Latent Causal Dynamic
- 18:25Causal Discovery from Nonstationary
- 19:34Nonlinear ICA with Multiple Domain
- 22:12With Changing Causal Relatior
- 22:59With Changing Causal Relations amon Latent Variables: Partial Identifiabilita
- 25:22Summary
related talks
32mNew Approaches To Learning Nonparametric (Latent) Causal Graphical Models
56mWhat can be passively learned about causality?
01h 02mUnmeasured Confounding and More Recent Developments/Challenges in Causal Discovery
24mA Friendly Introduction To Causality
29m🧠🧹 Causality - Mental Hygiene for Data Science
01h 06mIntroduction to Causal Graphical Models: Graphs, d-separation, do-calculus
44mCausal Reductions
01h 01mThe Mathematics of Causal Inference, with Reflections on Machine Learning and the Logic of Science