Reproduce “The Functional Neural Process” 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
Christos Louizos, Xiahan Shi, Klamer Schutte, Max Welling
Functional Neural Processes (FNPs) model uncertainty by learning links among data points, using related reference points to shape local predictions rather than a shared latent variable. FNP+ also conditions on each point’s embedding. On CIFAR-10, FNP+ achieves 7.2% test error and mean OOD AUCR of 94.9 ± 1.6, versus Neural Processes’ 7.5% and 90.1 ± 0.7 across five OOD datasets; it does not improve every case.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.