Reproduce “The Causal-Neural Connection: Expressiveness, Learnability, and Inference” 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
Kevin Xia, Kai-Zhan Lee, Yoshua Bengio, Elias Bareinboim
The authors introduce neural causal models (NCMs) that use a supplied causal graph to constrain neural mechanisms and noise. They test identifiability by comparing intervention effects across models that fit the same observations, then estimate effects when identified. In synthetic binary-variable experiments, the method distinguishes four identifiable from four non-identifiable cases; results do not establish real-world scalability or graph discovery.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.