Causal Discovery with Continuous Additive Noise Models
What this work does
Uses independence between causes and additive noise to identify causal direction under restrictions; RESIT searches for sinks by regression-residual independence and then removes unnecessary edges.
Connection to discovery
Demonstrates why observational discovery can work only after a mechanism-family assumption adds information beyond conditional independence.
Evidence
Identifiability propositions explicitly characterize exceptional families, including familiar linear-Gaussian ambiguity. Population arguments are distinguished from practical regression and independence tests.
Scope and limitations
Additive-noise restrictions, causal sufficiency, acyclicity and finite-sample test quality matter. Fitting an additive-noise model does not certify that the real mechanism belongs to that family.
THE RESEARCH CONVERSATION
Discussion & reactions
Share a question, an insight, or a replication experience. Sign in with GitHub to join the conversation.