Reproduce “Constraint-based Causal Discovery for Non-Linear Structural Causal Models with Cycles and Latent Confounders” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Patrick Forré, Joris M. Mooij
The authors introduce sigma-separation, an extension of d-separation that accounts for directed loops in nonlinear cyclic causal models. It gives conditional-independence implications for modular structural causal models, while discovery additionally requires sigma-faithfulness. Their algorithm uses observational and intervention data to rank graphs. In simulations, direct-edge AUC rose from 0.61 without interventions to 0.92 with five; this is synthetic evidence only.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.