Reproduce “Doubly Robust Proxy Causal Learning with Neural Mean Embeddings” 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
Bariscan Bozkurt, Alexandre Galashov, Dimitri Meunier, Zikai Shen, Arthur Gretton, Houssam Zenati
DR-PCL-Net estimates causal response functions from observational data using complementary proxy-based outcome and treatment bridges, without recovering the latent confounder. It learns adaptive neural features for both routes and combines their estimates, extending doubly robust proxy learning to heterogeneous and treatment-conditional responses, including structured treatments. On Demand at 10,000 observations, its causal MSE is 17.37 ± 1.07 versus OutcomeNet’s 22.42 ± 1.81; this comparison is limited to that benchmark and sample size.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.