Research map / Learning theory and algorithms / Matrix and tensor methods
Projection techniques to update the truncated SVD of evolving matrices with applications
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 |
|---|---|---|---|
| projection-based row update (Alg. 1, Z from Prop. 3) instead of Zha & Simon | truncated SVD updating for LSI · MED, CRAN, CISI precision (k=25,50,100) | average precision: e.g. MED k=25 59% vs 38% | Helped |
| projection-based row update instead of Zha & Simon for Eigenfaces | Eigenfaces PCA updating · Yale and AT&T classification | mean error rate: Yale k=50 27.56% vs 36.89% | Helped |
| projection-based row update instead of Zha & Simon | truncated SVD updating for LSI · Reuters wall-clock time | time | Helped |
| enhance projection matrix Z with X_lambda,r (randomized resolvent term) instead of Z=[Uk; Is] (ablation) | Alg. 1 projection matrix · MED sigma_50 triplet, r from 10 to 50 | residual norm: residual 0.053 at r=50 vs 0.269 without | Helped |
| Z built from Prop. 5 (with resolvent enhancement) instead of Prop. 3 (ablation) | Alg. 1 accuracy over updates · MED and CRAN singular triplets | accuracy deterioration over updates | Helped |
| Z from Prop. 5 (resolvent) instead of Prop. 3 (Reuters timing) (ablation) | Alg. 1 wall-clock time · Reuters | time | Hurt |
| thicker updates (fewer batches phi) vs many small updates | Alg. 1 relative to baselines · Reuters wall-clock time | time | Mixed |
Closest papers by meaning
- Fast Updating Truncated SVD for Representation Learning with Sparse Matrices (ICLR 2024 · similarity 0.91)
- Entrywise error bounds for low-rank approximations of kernel matrices (NeurIPS 2024 · similarity 0.82)
- Operator SVD with Neural Networks via Nested Low-Rank Approximation (ICML 2024 · similarity 0.82)
- The Structural Complexity of Matrix-Vector Multiplication (NeurIPS 2025 · similarity 0.82)
- Generalization Bound and Learning Methods for Data-Driven Projections in Linear Programming (NeurIPS 2024 · similarity 0.81)
- Fast exact recovery of noisy matrix from few entries: the infinity norm approach (NeurIPS 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.