Research map / Learning theory and algorithms / Matrix and tensor methods
Sparse Topic Modeling via Spectral Decomposition and Thresholding
Huy Tran, Yating Liu, Claire Donnat
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 TTS instead of Topic-SCORE | topic estimation · Research abstracts | average topic resolution: 0.145 -> 0.332 | Improved |
| use TTS instead of LDA | topic estimation · Research abstracts | average topic resolution: 0.304 -> 0.332 | Improved |
| add thresholding of infrequent words (alpha above 0) (ablation) | TTS · Synthetic pLSI, p=5000, n=N=500 | L1 error | Improved |
| raise the threshold excessively (ablation) | TTS · Synthetic pLSI, p=5000, n=N=500 | L1 error | Hurt |
| use Archetype Analysis instead of SP/SVS vertex hunting (ablation) | TTS · Synthetic pLSI without anchor words | L1 error | Improved |
| change anchor word frequency delta_anchor (ablation) | TTS · Synthetic pLSI | L1 error | No clear effect |
| increase Zipf exponent a_zipf (ablation) | TTS · Synthetic pLSI | L1 error | Improved |
| replace Zipf-decay columns of A with uniform non-anchor entries (ablation) | TTS vs Topic-SCORE · Synthetic pLSI (uniform generation) | L1 error | No clear effect |
| increase number of topics K to above 15 with n=1000 | TTS vs LDA/AWR · Synthetic pLSI, n=p=1000, N=500 | L1 error | Hurt |
Closest papers by meaning
- Optimal Estimation of Sparse Topic Models (JMLR 2020 · similarity 0.92)
- Probabilistic Non-Negative Matrix Factorization and Its Robust Extensions for Topic Modeling (AAAI 2017 · similarity 0.90)
- Spectral Learning for Supervised Topic Models (TPAMI 2017 · similarity 0.89)
- Provable Algorithms for Inference in Topic Models (ICML 2016 · similarity 0.88)
- Representing Mixtures of Word Embeddings with Mixtures of Topic Embeddings (ICLR 2022 · similarity 0.87)
- Diagnosing and Improving Topic Models by Analyzing Posterior Variability (AAAI 2018 · similarity 0.87)
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.
