AI Paper Atlas

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

Deep Autoencoding Topic Model With Scalable Hybrid Bayesian Inference

Hao Zhang, Bo Chen, Yulai Cong, Dandan Guo, Hongwei Liu, Mingyuan Zhou

TPAMI 2020 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Disentangled and discrete latent models · 32 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
use Weibull reparameterization instead of gamma approximation (RSVI) (ablation)DATM · 20News perplexityperplexity: 604 -> 581Improved
use stochastic-gradient MCMC (TLASGR) for global parameters instead of plain SGD (ablation)DATM-WAI · 20News perplexityperplexity: 581 -> 581 (20News); 984 -> 980 (Wiki)Improved
add stochastic downward path to inference network (ablation)DATM-IWHAI · 20News perplexityperplexity: 588 -> 581Improved
add depth: 128 to 128-64 to 128-64-32 layers (ablation)DATM-WHAI · 20News perplexityperplexity: 591 -> 582 -> 581Improved
add depth to DATM-IWHAI (no downward path) (ablation)DATM-IWHAI · 20News perplexityperplexity: 592 -> 589 -> 588No clear effect
replace logistic-normal AVITM latent with Weibull in single-layer modelAVITM · 20News perplexityperplexity: 654 -> 591Improved
learn layer widths via layer-wise training with K1max=128DATM-WHAI · 20News perplexityperplexity: 581 -> 578Improved
increase K1max from 64 to 1024 (ablation)DATM-WHAI · 20News perplexityperplexity: 584 -> 573Improved
use hybrid inference with TLASGR instead of Gibbs sampling at test timeDLDA-Gibbs · 20News test timeseconds per document: 10.46 -> 0.63Improved
add nonlinearity from latent features to labels (sDATM-N vs sDATM-L) (ablation)sDATM · 20News classificationerror rate (%): 18.63 -> 15.81Improved
add nonlinearity (sDATM-N vs sDATM-L) (ablation)sDATM · MNISTerror rate (%): 1.03 -> 0.97Improved
add label supervision to DATMDATM · IMDB classificationerror rate (%): 20.42 -> 10.92Improved

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