Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Bayesian Deep Embedding Topic Meta-Learner
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Adding the embedding-based topic meta-learner (global and task-specific topic embeddings) to DETM | Deep embedded topic model (DETM) · 20NG and RCV1 5-way few-shot classification | accuracy: Meta-DETM-3 45.5 vs HNS-DETM-3 42.5 (20NG 5-shot); 72.3 vs 66.6 (RCV1 5-shot) | Helped |
| Fine-tuning and MAML-style meta-learning of all DETM parameters (MAML-DETM) (ablation) | DETM · 20NG per-heldout-word perplexity | perplexity | Hurt |
| Increasing the number of layers from 1 to 3 in Meta-DETM (ablation) | Meta-DETM · 5-way few-shot classification on 20NG and RCV1 | accuracy: 20NG 5-shot 44.1 (1-layer) vs 45.5 (3-layer); RCV1 5-shot 66.4 vs 72.3 | Helped |
| Using a task-specific decoder instead of a shared decoder for each task | Bi-level generative model · few-shot topic modeling | perplexity | No clear effect |
| Fine-tuning pre-trained traditional topic models (LDA, GBN) on the support set | LDA and GBN · per-heldout-word perplexity on 20NG, Yahoo, RCV1 | perplexity | Hurt |
| Tuning the regularization coefficients gamma and lambda across layers (ablation) | Meta-DETM · 20NG test perplexity | perplexity | Mixed |
Closest papers by meaning
- Knowledge-Aware Bayesian Deep Topic Model (NeurIPS 2022 · similarity 0.91)
- TopicNet: Semantic Graph-Guided Topic Discovery (NeurIPS 2021 · similarity 0.89)
- Sawtooth Factorial Topic Embeddings Guided Gamma Belief Network (ICML 2021 · similarity 0.89)
- Topic-VQ-VAE: Leveraging Latent Codebooks for Flexible Topic-Guided Document Generation (AAAI 2024 · similarity 0.86)
- FASTopic: Pretrained Transformer is a Fast, Adaptive, Stable, and Transferable Topic Model (NeurIPS 2024 · similarity 0.85)
- On the Affinity, Rationality, and Diversity of Hierarchical Topic Modeling (AAAI 2024 · similarity 0.85)
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.