Research map / Learning theory and algorithms / Prediction and decision losses
Sign-Full Random Projections
Ping Li
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 the normalized sign-split estimator instead of sign-sign projections | Near-neighbor ranking · RCV1, ρ₀=0.95 | precision-recall | Improved |
| Use the normalized sign-split estimator instead of sign-sign projections | Near-neighbor ranking · RCV1, ρ₀≤0.6 | precision-recall | Improved |
| Normalize the query projections y (ablation) | Sign-full signed-product estimator ρ̂_g · Cosine similarity estimation | asymptotic variance: V_{g,n} ≤ V_g | Improved |
| Normalize the query projections y (ablation) | Sign-full sign-split estimator ρ̂_s · Cosine similarity estimation, ρ≤(√3−1)/2 | asymptotic variance: V_s≥V_{s,n} | Improved |
| Normalize the query projections y (ablation) | Sign-full sign-split estimator ρ̂_s · Cosine similarity estimation, ρ>0.3660 | asymptotic variance: 0.3660 | Hurt |
| Use the sign-split estimator rather than the signed-product estimator (ablation) | Sign-full normalized estimator · High similarity, ρ→1 | asymptotic variance ratio to sign-sign: 4/(3π)≈0.4244; ∞ | Improved |
| Use the sign-split estimator rather than the signed-product estimator (ablation) | Sign-full normalized estimator · Near-neighbor search, high similarity | precision-recall | Improved |
| Normalize y in the sign-split estimator (ablation) | Sign-full estimator ρ̂_s · Cosine similarity estimation, ρ→1 | asymptotic variance ratio to sign-sign: 4/(3π)≈0.4244 | No clear effect |
| Use normalized sign-split estimator instead of the sign-full MLE | Sign-full cosine similarity estimation · ρ≥−0.4 | asymptotic variance: V_{s,n}/V_m ≈ 1.18 | Hurt |
| Use normalized sign-split estimation rather than sign-sign estimation | Near-neighbor search · ρ close to 1, k=10 | MSE improvement factor: as much as a factor of 8 | Improved |
Closest papers by meaning
- Improving Sign Random Projections With Additional Information (ICML 2018 · similarity 0.90)
- Random Projections with Asymmetric Quantization (NeurIPS 2019 · similarity 0.88)
- Node Similarities under Random Projections: Limits and Pathological Cases (ICLR 2025 · similarity 0.84)
- MaxSketch: Robust Distinct Counting in Streams via Random Projections (NeurIPS 2026 · similarity 0.83)
- When and Why SignSGD Outperforms SGD: A Theoretical Study Based on $\ell_1$-norm Lower Bounds (NeurIPS 2026 · similarity 0.82)
- Noisy SIGNSGD Is More Differentially Private Than You (Might) Think (ICML 2025 · similarity 0.81)
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.
