Reproduce “A Deep and Tractable Density Estimator” 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

Benigno Uria, Iain Murray, Hugo Larochelle

The authors introduce an order-agnostic neural autoregressive distribution estimator, sharing parameters across variable orderings so users can select an ordering suited to a query. Input masks distinguish observed zeros from omitted variables. On binarized MNIST, a 128-order, two-layer ensemble scores −85.10, better than fixed-order NADE at −88.33 but below estimated DBN at −84.55; exact inference applies only to suitably arranged queries.

Summary by alphaXiv.

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