AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Document Neural Autoregressive Distribution Estimation

Stanislas Lauly, Yin Zheng, Alexandre Allauzen, Hugo Larochelle

JMLR 2017 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Tractable probabilistic models · 16 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add hidden layers to DeepDocNADE (1 to 3 layers) (ablation)DeepDocNADE · 20 Newsgroups (50 topics)test perplexity per word: 835 -> 877 -> 923Hurt
add hidden layers to DeepDocNADE (1 to 3 layers) (ablation)DeepDocNADE · RCV1-V2 (50 topics)test perplexity per word: 579 -> 552 -> 539Improved
replace DocNADE with one-hidden-layer DeepDocNADEDocNADE · 20 Newsgroups (50 topics)test perplexity per word: 896 -> 835Improved
replace DocNADE with one-hidden-layer DeepDocNADEDocNADE · RCV1-V2 (50 topics)test perplexity per word: 742 -> 579Improved
increase number of word-ordering ensembles M (ablation)DeepDocNADE · 20 Newsgroups and RCV1-V2test perplexity per wordNo clear effect
add DocNADE context component to feed-forward language model (ablation)FFN · APNews language modeling (100 topics)test perplexity per word: 119.78 -> 111.93Improved
increase number of grouped sentences per document from 1 to 5 (ablation)DocNADE-LM · APNews language modeling (100 topics)test perplexity per word: 111.93 -> 109.22Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy