Reproduce “Parametric Rates for Nonparametric Conditional-Moment Bilevel Optimization” 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

Fares El Khoury, Houssam Zenati, Nathan Kallus, Michael Arbel, Aurélien Bibaut

OBiGrad corrects hypergradient estimates in conditional-moment bilevel optimization by adding a residual term based on regressions for the inner prediction, its sensitivity, and the conditional outcome mean, canceling first-order nuisance-error effects. In a synthetic zero-residual instrumental-variable design at N=800, Gaussian Nadaraya–Watson root RMSE fell from 0.0251 to 0.0124; this measures stationary-point estimation, not outer-loop convergence.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.