Research map / Probabilistic and scientific modeling / Variational and generative sampling
Choosing the Number of Topics in LDA Models – A Monte Carlo Comparison of Selection Criteria
Victor Bystrov, Viktoriia Naboka-Krell, Anna Staszewska-Bystrova, Peter Winker
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 sBIC instead of Cao_Juan, Mimno or OpTop | topic number selection · DGP1 (K_true=38) | F1: 0.97 vs 0.83, 0.85, 0.81 | Improved |
| use sBIC instead of other criteria | topic number selection · DGP2 (K_true=12) | F1: 0.92 vs 0.87, 0.29, 0.55 | Improved |
| use sBIC instead of Cao_Juan | topic number selection · DGP3 (K_true=70) | F1: 0.91 vs 0.91 | No clear effect |
| use sBIC instead of OpTop | topic number selection · DGP1/2/3 | mean estimated topics: OpTop 5%: 55.81, 31.70, 88.03 vs sBIC 40.15, 12.28, 65.43 | Improved |
| use sBIC instead of Mimno on small corpus | topic number selection · DGP2 (K_true=12) | recall: 0.17 vs 0.93 | Improved |
| change OpTop low-frequency word cut-off from 5% to 20% (ablation) | OpTop · DGP1/2/3 | F1: 0.81, 0.55, 0.87 for both | No clear effect |
| sBIC on DGP1 versus true topic count | sBIC · DGP1 (K_true=38) | mean estimated topics: 40.15 vs 38 | Hurt |
| sBIC on DGP3 versus true topic count | sBIC · DGP3 (K_true=70) | mean estimated topics: 65.43 vs 70 | Hurt |
| use cosine similarity instead of binary match for recall/precision (ablation) | evaluation · DGP1/2/3 | F1 | No clear effect |
Closest papers by meaning
- A Statistical Approach for Optimal Topic Model Identification (JMLR 2022 · similarity 0.88)
- Principled Selection of Hyperparameters in the Latent Dirichlet Allocation Model (JMLR 2018 · similarity 0.85)
- E-LDA: Toward Interpretable LDA Topic Models with Strong Guarantees in Logarithmic Parallel Time (ICML 2025 · similarity 0.84)
- Online Bayesian Moment Matching for Topic Modeling with Unknown Number of Topics (NeurIPS 2016 · similarity 0.84)
- A Reduction for Efficient LDA Topic Reconstruction (NeurIPS 2018 · similarity 0.84)
- Discriminative Topic Modeling with Logistic LDA (NeurIPS 2019 · similarity 0.83)
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.
