Reproduce “Learning Conditional Expectation Operators via Functional Newton Updates” 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
Thiago Ramos, Alek Fröhlich, Daniel Perazzo, Massimiliano Pontil
FSNM learns a low-rank representation of dependence between paired variables, enabling multiple conditional queries from one fit. It alternates preconditioned conditional-mean regressions, approximated with regression trees, to estimate a centered density-ratio kernel. In synthetic tests it recovers the broad pattern but retains tree-shaped residuals; on a restricted glass test, its exceedance-probability query reduces Brier score 67.6% versus climatology.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.