Seminar on Learning Causal Models
Description
Schedule
Seminar 1 and 2 introduce connections between causal interpretations of graphs and their conditional independence properties. This seminar will discuss how these connections can be applied to the problem of learning about causal relations from data.
We consider both Bayesian and asymptotic approaches, with an emphasis on the former. We relate causal interpretations to commonly used assumptions used for the selection of graph structure such as parameter independence, parameter modularity, and marginal likelihood equivalence. In addition, we address difficulties in scoring and searching over graphical models with latent variables, compare model selection to model averaging techniques, and discuss assumptions under which "counterfactual" information can be learned.
INVITED SPEAKERS
| G. Cooper University of Pittsburgh | J. Andersen Aalborg University |
| B. Frey University of Waterloo | J. Cheng University of Alberta |
| T. Richardson University of Warwick | G. Shafer Rutgers University |
| P. Giudici University of Pavia | J. Whittaker Lancaster University |
| R. Shachter Stanford University |
