Score-Based Learning of Cluster DAGs from Interventions

Name: Score-Based Learning of Cluster DAGs from Interventions
Uploaded: 2026-10-14T15:15:55.403576+00:00
Duration: 280 s
Description: The paper introduces COARSE, a score-based method to learn an interventional coarsening from heterogeneous data. It merges variables with identical intervention signatures and uses marginal shifts to identify clusters and their causal order. This approach replaces constraint-based edge recovery with a faster, single local search per cluster.

The paper introduces COARSE, a score-based method to learn an interventional coarsening from heterogeneous data. It merges variables with identical intervention signatures and uses marginal shifts to identify clusters and their causal order. This approach replaces constraint-based edge recovery with a faster, single local search per cluster.

Open the full episode