Research map / Label-efficient and robust learning / Recognition and long-tail
Fine-Grained Change Point Detection for Topic Modeling with Pitman-Yor Process
Feifei Wang, Zimeng Zhao, Ruimin Ye, Xiaoge Gu, Xiaoling Lu
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 incorrect PYP hyperparameters a=0.1, b=5 in estimation (ablation) | TOPIC-PYP · Synthetic Scenario 2, true a=0.5 b=5 | accuracy: 0.99 -> 0.96 | No clear effect |
| use incorrect PYP hyperparameters a=0.9, b=5 in estimation (ablation) | TOPIC-PYP · Synthetic Scenario 2, true a=0.5 b=5 | accuracy: 0.99 -> 0.98 | No clear effect |
| use incorrect PYP hyperparameter b=100 in estimation (ablation) | TOPIC-PYP · Synthetic Scenario 2, true a=0.5 b=5 | accuracy: 0.99 -> 0.99 | No clear effect |
| unify topic modeling and change point detection (TOPIC-PYP vs two-stage methods) | topic change point detection · Synthetic Scenario 1-4 | coherence score: 29.065 vs 3.826 (Rolling LDA), Scenario 1 | Improved |
| replace Topic-CD with TOPIC-PYP | unified change point model · Synthetic Scenario 2 | P score: 0.318 -> 0.001 | Improved |
| replace Topic-CD with TOPIC-PYP | unified change point model · Synthetic Scenario 4 (h=0) | precision: 0.625 -> 0.709 | Improved |
| increase the period T (ablation) | TOPIC-PYP · Synthetic, T in 10,20,30 and K in 5,10,15 | precision: precision decreases as T increases; recall stays above 90% | Hurt |
| increase PYP hyperparameter a (b=5) (ablation) | TOPIC-PYP · Synthetic Scenario 2 | precision and recall: precision and accuracy increase, recall decreases | Mixed |
| increase PYP hyperparameter b (a=0.5) (ablation) | TOPIC-PYP · Synthetic Scenario 2 | precision and recall: precision increases, recall decreases, accuracy nearly unchanged | Mixed |
| vary lambda, alpha and gamma (ablation) | TOPIC-PYP · Synthetic Scenario 2 | change point detection performance | No clear effect |
| use unified model instead of two-stage methods | topic change point detection · Synthetic runtime | running time | Hurt |
Closest papers by meaning
- Neural Dynamic Focused Topic Model (AAAI 2023 · similarity 0.84)
- IMS-DTM: Incremental Multi-Scale Dynamic Topic Models (AAAI 2018 · similarity 0.84)
- Scalable inference of topic evolution via models for latent geometric structures (NeurIPS 2019 · similarity 0.83)
- ChangeDiff: A Multi-Temporal Change Detection Data Generator with Flexible Text Prompts via Diffusion Model (AAAI 2025 · similarity 0.83)
- UniChange: Unifying Change Detection with Multimodal Large Language Model (CVPR 2026 · similarity 0.81)
- Provable unlearning in topic modeling and downstream tasks (ICLR 2025 · similarity 0.80)
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.
