Reproduce “Deep AutoRegressive Networks” 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
Karol Gregor, Ivo Danihelka, Andriy Mnih, Charles Blundell, Daan Wierstra
DARN generates observations by drawing stochastic hidden units in sequence, with each unit conditioned on earlier units, then decoding the resulting representation. The authors train the encoder and decoder jointly by minimizing an approximate minimum-description-length cost. On binarized MNIST, a 500-unit model reports 84.13 ± 0.01 nats per example, below EoNADE’s 85.10; this is an importance-sampling estimate, not an exact evaluation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.