Research map / Learning theory and algorithms / Matrix and tensor methods
Zero-Inflated Exponential Family 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 |
|---|---|---|---|
| add zero-inflation (ZIE-0) to EFE (ablation) | EFE · eBird K=64, all observations | predictive log-likelihood: −0.374 -> −0.308 | Helped |
| add exposure covariates (ZIE-cov) over ZIE-0 (ablation) | ZIE-0 · eBird K=64, all observations | predictive log-likelihood: −0.308 -> −0.298 | Helped |
| ZIE-0 vs downweighting zeros (EFE-dz) on positives (ablation) | EFE-dz · eBird K=64, positive | predictive log-likelihood: −1.727 -> −1.736 | No clear effect |
| downweight zeros (EFE-dz) vs EFE (ablation) | EFE · eBird K=64, all observations | predictive log-likelihood: −0.374 -> −0.490 | Hurt |
| downweight zeros (EFE-dz) vs EFE on positives (ablation) | EFE · eBird K=64, positive | predictive log-likelihood: −2.140 -> −1.727 | Helped |
| other negative weights 0.05, 0.2 for EFE-dz (ablation) | EFE-dz · eBird | predictive log-likelihood | Mixed |
| zero-inflated exposure vs EFE under down-sampling | ZIE · eBird K=64, d.s. ratio 0.005 | delta p: 0.870 -> 0.463 | Helped |
| ZIE-cov vs EFE | EFE · Movie ratings K=32, all | predictive log-likelihood: −0.450 -> −0.349 | Helped |
| ZIE-0 vs EFE-dz on positives | EFE-dz · Movie ratings K=32, positive | predictive log-likelihood: −1.152 -> −1.267 | Hurt |
| add informative exposure covariates (ZIE-cov) vs ZIE-0 (ablation) | ZIE-0 · Market K=64, positive | predictive log-likelihood: −2.190 -> −1.127 | Helped |
| ZIE-cov vs EFE | EFE · Market K=64, positive | predictive log-likelihood: −4.581 -> −1.127 | Helped |
| use POS tags for exposure (ZIE-cov) vs EFE (ablation) | EFE · Wikipedia K=64 | log-likelihood per document: −51.1 -> −43.1 | Helped |
Closest papers by meaning
- Exponential Family Embeddings (NeurIPS 2016 · similarity 0.89)
- Structured Embedding Models for Grouped Data (NeurIPS 2017 · similarity 0.87)
- Provable Gaussian Embedding with One Observation (NeurIPS 2018 · similarity 0.87)
- Zero-Shot Robustification of Zero-Shot Models (ICLR 2024 · similarity 0.83)
- NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Models (ICLR 2025 · similarity 0.83)
- ZeroDiff: Solidified Visual-semantic Correlation in Zero-Shot Learning (ICLR 2025 · 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.