Research map / Learning theory and algorithms / Matrix and tensor methods
A Fast Frequent Directions Algorithm for Low Rank Approximation
Dan Teng, Delin Chu
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 SpFD instead of SpEmb alone | low-rank approximation · synthetic and real datasets | F-norm and 2-norm relative error | Improved |
| use SpFD instead of exact FD | low-rank approximation · synthetic and real datasets | running time | Improved |
| use SpFD instead of FD for accuracy | low-rank approximation · some real datasets | relative error | Mixed |
| increase q from 5 to 50 in SpFD (ablation) | SpFD · real datasets | relative error | Improved |
| increase q from 5 to 50 in SpFD (ablation) | SpFD · Computational Complexity network | running time (s): 0.0079 -> 0.0062 | Improved |
| increase q from 5 to 50 in SpFD (ablation) | SpFD · Death Penalty network | running time (s): 0.0213 -> 0.0188 | Improved |
| increase q from 5 to 50 in SpFD (ablation) | SpFD · Death Penalty network, top-10 authorities | nodes shared with e^A: 8 -> 10 | Improved |
| use SpFD50 instead of SpEmb | hub/authority scoring · Computational Complexity network, top-10 authorities | nodes shared with e^A: 5 -> 8 | Improved |
| use SpFD50 instead of FD | hub/authority scoring · Computational Complexity network | running time (s): 0.0814 -> 0.0062 | Improved |
| use low-rank approximation instead of exact e^A | hub/authority scoring · Death Penalty network | running time (s): 5.1974 -> 0.0188 | Improved |
| use e^A instead of HITS | hub/authority ranking · Computational Complexity network | overlap with HITS top 10: 4 of 10 shared hubs and authorities | Mixed |
| use NormSamp instead of FD | hub/authority scoring · Computational Complexity network, top-10 authorities | nodes shared with e^A: 9 -> 3 | Hurt |
Closest papers by meaning
- Efficient frequent directions algorithms for approximate decomposition of matrices and higher-order tensors (JMLR 2026 · similarity 0.90)
- Few-Shot Data-Driven Algorithms for Low Rank Approximation (NeurIPS 2021 · similarity 0.87)
- Learning-Based Low-Rank Approximations (NeurIPS 2019 · similarity 0.86)
- Consistent Low-Rank Approximation (ICLR 2026 · similarity 0.86)
- Subspace Clustering With Priors via Sparse Quadratically Constrained Quadratic Programming (CVPR 2016 · similarity 0.85)
- Random features for Grassmannian kernel approximation with bounded rank-one projections (NeurIPS 2026 · similarity 0.84)
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.
