Research map / Learning theory and algorithms / Matrix and tensor methods
Subspace Embedding and Linear Regression with Orlicz Norm
Alexandr Andoni, Chengyu Lin, Ying Sheng, Peilin Zhong, Ruiqi Zhong
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 normalized Huber Orlicz loss instead of ℓ1 or ℓ2 loss | Simulated regression · Mixed noise, balance and overconstraint | ℓ2 distance to ground truth: 37.9/37.8/37.5; 8.7/7.6/7.5 | Improved |
| Use normalized Huber Orlicz loss instead of ℓ2 loss | Simulated regression · Gaussian noise, balance | ℓ2 distance to ground truth: 194.5 -> 197.3 | Hurt |
| Use normalized Huber Orlicz loss instead of ℓ1 loss | Simulated regression · Sparse noise, balance and overconstraint | ℓ2 distance to ground truth: 25.3 -> 30.0; 2e-9 -> 1.5 | Hurt |
| Use normalized Huber Orlicz loss instead of ℓ2 loss | Simulated regression · Gaussian noise, overconstraint | ℓ2 distance to ground truth: 20.0 -> 24.9 | Hurt |
| Use normalized Huber Orlicz loss instead of ℓ1 loss | Synthetic ℓ1 rank-5 approximation · Synthetic data | ℓ1 loss (× 10^4): 1.36 -> 1.04 | Improved |
| Use exponential-random-variable Orlicz regression with G_ℓ1.5 | Simulated regression · Gaussian noise and one sparse outlier (scale 100) | ℓ2 distance to ground truth: 14.7 | Improved |
| Decrease Huber δ as sparse-noise scale increases (ablation) | Orlicz regression with normalized Huber loss · Mixed Gaussian and sparse noise, scales s from [0, 0.5, 1, 2] | ℓ2 distance to ground truth | Improved |
Closest papers by meaning
- Efficient Symmetric Norm Regression via Linear Sketching (NeurIPS 2019 · similarity 0.89)
- Sparse Dimensionality Reduction Revisited (ICML 2024 · similarity 0.86)
- Subspace Fitting Meets Regression: The Effects of Supervision and Orthonormality Constraints on Double Descent of Generalization Errors (ICML 2020 · similarity 0.86)
- Beyond Worst-Case Dimensionality Reduction for Sparse Vectors (ICLR 2025 · similarity 0.84)
- Efficient Alternating Minimization with Applications to Weighted Low Rank Approximation (ICLR 2025 · similarity 0.84)
- Minimum Norm Interpolation Meets The Local Theory of Banach Spaces (ICML 2024 · 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.
