Research map / Learning theory and algorithms / Matrix and tensor methods
Fast and Accurate Least-Mean-Squares Solvers for High Dimensional Data
Alaa Maalouf, Ibrahim Jubran, Dan Feldman
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 |
|---|---|---|---|
| run LMS solver on the Caratheodory coreset instead of full data | sklearn LMS solvers with cross validation · Synthetic and UCI datasets | running time: x10--x100 | Improved |
| run linear regression on coreset instead of full data | LinearRegression · Synthetic data, various d | running time: up to x10 | Improved |
| use Intel's Python distribution instead of CPython | boosted and unboosted solvers · Synthetic data | running time: 15-40 | Improved |
| use sparse Algorithm instead of Caratheodory-matrix variant (k1,k2 tuned) | LMS-Coreset++ · Synthetic data | running time | Improved |
| set k2 to (d+1)^2 divided by 12 rather than the theoretical maximum (ablation) | LMS-Coreset++ · Synthetic data | running time: 12 | Improved |
| use coreset instead of SKETCH+CHOLESKY in 32-bit float | LMS solvers · Datasets (i) and (iii) | numerical accuracy: up to orders of magnitude | Improved |
| use coreset instead of SKETCH+CHOLESKY in 32-bit float | LMS solvers · Datasets (i) and (iii) | running time: slower by at most a factor of 2 | Hurt |
| use coreset instead of SKETCH+INVERSE | LinearRegression · Datasets (i),(ii) | numerical accuracy | Improved |
| use coreset for small n | LMS-Coreset · small data size n | running time | Hurt |
Closest papers by meaning
- Fast and Accurate Least-Mean-Squares Solvers (NeurIPS 2019 · similarity 0.98)
- Householder Sketch for Accurate and Accelerated Least-Mean-Squares Solvers (ICML 2021 · similarity 0.88)
- Solving L1-regularized SVMs and Related Linear Programs: Revisiting the Effectiveness of Column and Constraint Generation (JMLR 2022 · similarity 0.87)
- Faster Algorithms for Structured Linear and Kernel Support Vector Machines (ICLR 2025 · similarity 0.86)
- Precise asymptotics of reweighted least-squares algorithms for linear diagonal networks (NeurIPS 2024 · similarity 0.85)
- Distributed Least Squares in Small Space via Sketching and Bias Reduction (NeurIPS 2024 · similarity 0.85)
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.
