Research map / Probabilistic and scientific modeling / Variational and generative sampling
Dense Distributions from Sparse Samples: Improved Gibbs Sampling Parameter Estimators for LDA
Yannis Papanikolaou, James R. Foulds, Timothy N. Rubin, Grigorios Tsoumakas
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use dense soft-count theta^p estimator instead of hard-count theta with a single burned-in sample | CGS LDA · Unsupervised LDA held-out perplexity (BioASQ, NYT, Reuters, TASA, K=100) | perplexity | Improved |
| use phi^p estimator instead of standard phi | CGS LDA · Unsupervised LDA held-out perplexity (K=100) | perplexity | Improved |
| average over more samples/chains | theta vs theta^p · Unsupervised LDA held-out perplexity (K=100) | perplexity | No clear effect |
| use phi^p instead of phi for word association ranking | LDA trained on TASA · Word association norms (TASA) | median rank difference | Improved |
| use CGS_p estimators instead of CVB0 as number of topics grows | CGS_p · Unsupervised LDA perplexity vs topics | perplexity | Mixed |
| use VB instead of CVB0/CGS | LDA inference · Unsupervised LDA perplexity vs topics | perplexity | Hurt |
Closest papers by meaning
- Principled Selection of Hyperparameters in the Latent Dirichlet Allocation Model (JMLR 2018 · similarity 0.87)
- Pólya Urn Latent Dirichlet Allocation: A Doubly Sparse Massively Parallel Sampler (TPAMI 2018 · similarity 0.87)
- SoftCVI: Contrastive variational inference with self-generated soft labels (ICLR 2025 · similarity 0.87)
- On the properties of variational approximations of Gibbs posteriors (JMLR 2016 · similarity 0.86)
- Efficient Mixture Learning in Black-Box Variational Inference (ICML 2024 · similarity 0.85)
- Weighted Conformal Clustering (NeurIPS 2026 · similarity 0.84)
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.
