Research map / Learning theory and algorithms / Prediction and decision losses
Improving Sign Random Projections With Additional Information
Keegan Kang, Weipin Wong
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 |
|---|---|---|---|
| add the margin-based estimator to SRP | SRP · MNIST, k=1024 | average RMSE: 0.0471 -> 0.0451 | Improved |
| add the margin-based estimator to SRP | SRP · Gisette, k=1024 | average RMSE: 0.0466 -> 0.0419 | Improved |
| add the margin-based estimator to SBLSH | SBLSH · MNIST, k=1024 | average RMSE: 0.0386 -> 0.0378 | Improved |
| add the margin-based estimator to SBLSH | SBLSH · Gisette, k=1024 | average RMSE: 0.0446 -> 0.0405 | Improved |
| add the estimator to SRP: spread of RMSE across simulations | SRP · Gisette, k=1024 | average RMSE: 0.0091 -> 0.0025 | Improved |
| add the estimator to SBLSH: spread of RMSE across simulations | SBLSH · Gisette, k=1024 | average RMSE: 0.0079 -> 0.0021 | Improved |
| add the estimator to SRP: spread of RMSE across simulations | SRP · MNIST, k=1024 | average RMSE: 0.0055 -> 0.0049 | Improved |
| use SBLSH instead of SRP-est | SRP-est · MNIST and Gisette, large k | average RMSE: 0.0451 -> 0.0386 | Mixed |
| use SRP-est instead of SBLSH when pre-processing cost matters | SBLSH · MNIST and Gisette | average RMSE | No clear effect |
| use small k (below about 256 bits) | margin-based estimator · MNIST and Gisette, small k | average RMSE | Hurt |
| use large k (around 256 or more) | margin-based estimator · MNIST and Gisette | average RMSE | Improved |
| use two extra vectors e1 and e2 instead of one | margin-based estimator · MNIST and Gisette | average RMSE | Mixed |
Closest papers by meaning
- Sign-Full Random Projections (AAAI 2019 · similarity 0.90)
- Smooth Flipping Probability for Differential Private Sign Random Projection Methods (NeurIPS 2023 · similarity 0.85)
- Node Similarities under Random Projections: Limits and Pathological Cases (ICLR 2025 · similarity 0.81)
- When and Why SignSGD Outperforms SGD: A Theoretical Study Based on $\ell_1$-norm Lower Bounds (NeurIPS 2026 · similarity 0.81)
- Fundamental Bias in Inverting Random Sampling Matrices with Application to Sub-sampled Newton (ICML 2025 · similarity 0.80)
- Euclidean distance compression via deep random features (NeurIPS 2024 · similarity 0.80)
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.
