Research map / Learning theory and algorithms / Matrix and tensor methods
Streaming Principal Component Analysis From Incomplete Data
Armin Eftekhari, Gregory Ongie, Laura Balzano, Michael B. Wakin
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 |
|---|---|---|---|
| process overlapping blocks (SNIPE-overlap) instead of disjoint blocks | SNIPE · Synthetic, p=0.60, error < 10^-3 | columns revealed to reach error: 171.4 -> 71.3 | Improved |
| process overlapping blocks (SNIPE-overlap) instead of disjoint blocks | SNIPE · Synthetic, p=0.15, error < 10^-3 | columns revealed to reach error: 1815.7 -> 1588.3 | Improved |
| use SNIPE-overlap instead of GROUSE at low sampling | GROUSE · Synthetic, p=0.15, error < 10^-3 | columns revealed to reach error: 878.0 -> 1588.3 | Hurt |
| use SNIPE-overlap instead of GROUSE at high sampling | GROUSE · Synthetic, p=0.75, error < 10^-7 | columns revealed to reach error: 213.5 -> 91.8 | Improved |
| use SNIPE instead of PETRELS at high sampling | PETRELS · Synthetic, p=0.75, error < 10^-7 | columns revealed to reach error: 599.1 -> 241.6 | Improved |
| use a coherent subspace (diagonal scaling D[i,i]=1/i) instead of a generic one (ablation) | SNIPE · Synthetic, n=300, r=10, p=3r/n | estimation error d_G: 2.795e-5 -> 0.4286 | Hurt |
| increase block size b beyond 2r with fixed scope T (ablation) | SNIPE · Synthetic, n=100, r=5, p=3r/n | estimation error d_G | Hurt |
| increase ambient dimension n with p=3r/n (ablation) | SNIPE · Synthetic, r=5 | estimation error d_G | Hurt |
Closest papers by meaning
- Robust Subspace Approximation in a Stream (NeurIPS 2018 · similarity 0.87)
- MOSES: A Streaming Algorithm for Linear Dimensionality Reduction (TPAMI 2019 · similarity 0.86)
- Compact Matrix Factorization With Dependent Subspaces (CVPR 2017 · similarity 0.86)
- High-Rank Matrix Completion and Clustering under Self-Expressive Models (NeurIPS 2016 · similarity 0.86)
- High-Dimensional Geometric Streaming for Nearly Low Rank Data (ICML 2024 · similarity 0.85)
- Oja's Algorithm for Streaming Sparse PCA (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.
