Research map / Label-efficient and robust learning / Recognition and long-tail
Group-Sparse Matrix Factorization for Transfer Learning of Word Embeddings
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 |
|---|---|---|---|
| group-sparse (l2,1) penalty joint estimator vs gold-only and proxy-only estimators | word embeddings with synthetic low-data setting (ng=50, np=5000, d=20) · synthetic data | Frobenius error of Theta_g: error ~4% of proxy and ~2% of gold | Helped |
| group-sparse penalty on GloVe (joint estimator) vs Mittens (l2 penalty) | GloVe domain-word identification · Wikipedia domains (Table 2) | weighted F1-score: Joint beats Mittens in all four domains (e.g., finance 0.2280 vs 0.1912) | Helped |
| group-sparse penalty vs Mittens and CCA/KCCA for domain word identification | single-article Wikipedia finance domain, varying top percentage 10-30% · Wikipedia finance (Figure 2) | weighted F1-score: consistently higher than baselines across thresholds | Helped |
| group-sparse penalty on GloVe vs Mittens | clinical trial eligibility prediction with GloVe embeddings (50 observations) · clinical trial eligibility | F1-score: 0.612554 vs 0.604338 | Helped |
| group-sparse penalty vs Mittens on Word2Vec and Dict2Vec embeddings | clinical trial eligibility prediction · clinical trial eligibility | F1-score: Word2Vec 0.624428 vs 0.604981; Dict2Vec 0.638923 vs 0.627772 | Helped |
| group-sparse penalty on Dict2Vec vs proxy-only embeddings (ablation) | clinical trial eligibility with Dict2Vec · clinical trial eligibility | F1-score: 0.638923 vs 0.623337 (proxy only) | Helped |
Closest papers by meaning
- A Compressed Sensing View of Unsupervised Text Embeddings, Bag-of-n-Grams, and LSTMs (ICLR 2018 · similarity 0.88)
- Understanding the Semantic Content of Sparse Word Embeddings Using a Commonsense Knowledge Base (AAAI 2020 · similarity 0.86)
- Transferring Linear Features Across Language Models With Model Stitching (NeurIPS 2025 · similarity 0.85)
- Beyond Matryoshka: Revisiting Sparse Coding for Adaptive Representation (ICML 2025 · similarity 0.84)
- Learning and Transferring Sparse Contextual Bigrams with Linear Transformers (NeurIPS 2024 · similarity 0.84)
- SASA: Subspace-Aware Sparse Autoencoders for Effective Mechanistic Interpretability (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.