Reproduce “Bayesian causal discovery from unknown general interventions” 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
Alessandro Mascaro, Federico Castelletti
This paper presents a Bayesian framework for causal discovery from interventional data, uniquely capable of handling interventions with unknown targets and complex effects. The approach demonstrated improved accuracy in learning causal structures on synthetic data and identified subtype-specific protein interaction changes in human cancer cell data.
Key insights
- A formal characterization of general interventions and I-Markov equivalence classes was provided, outlining the fundamental limits of causal identifiability.
- Two DAGs are I-Markov equivalent under general interventions if they share the same skeleton, v-structures, and compelled edges.
- A Bayesian approach with specific parameter priors was developed to ensure "score equivalence" for I-Markov equivalent DAGs, leading to robust inference.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.