Research map / Learning theory and algorithms / Clustering
Self-Representation Based Unsupervised Exemplar Selection in a Union of Subspaces
Chong You, Chi Li, Daniel P. Robinson, René Vidal
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 |
|---|---|---|---|
| add FFS exemplar selection instead of random exemplars (ESC-FFS vs ESC-Rand) (ablation) | ESC · GTSRB (14 classes, subspace clustering) | clustering accuracy (%): 89.7 -> 93.0 | Improved |
| add FFS exemplar selection instead of random exemplars (ablation) | ESC · GTSRB (14 classes, subspace clustering) | F-score (%): 75.5 -> 85.3 | Improved |
| use exemplars selected by FFS instead of all points (ESC-FFS vs SSC) | SSC · GTSRB (14 classes, subspace clustering) | clustering accuracy (%): 92.4 -> 93.0 | Improved |
| add FFS exemplar selection instead of random exemplars (ablation) | ESC · Extended Yale B (10 imbalanced classes, subspace clustering) | clustering accuracy (%): 65.7 -> 71.1 | Improved |
| replace SSC with ESC-FFS on imbalanced data | SSC · Extended Yale B (10 imbalanced classes, subspace clustering) | clustering accuracy (%): 67.1 -> 71.1 | Improved |
| use k-centers exemplar selection instead of random (ablation) | ESC · Extended Yale B (10 imbalanced classes, subspace clustering) | clustering accuracy (%): 65.7 -> 67.0 | No clear effect |
| use K-medoids exemplar selection instead of random (ablation) | ESC · Extended Yale B (10 imbalanced classes, subspace clustering) | clustering accuracy (%): 65.7 -> 64.5 | No clear effect |
| set t=3 instead of t=2 in tNN graph (ablation) | ESC-FFS · Extended Yale B (10 imbalanced classes, subspace clustering) | clustering accuracy (%): 61.2 -> 70.9 | Improved |
| set t=3 instead of t=8 in tNN graph (ablation) | ESC-FFS · Extended Yale B (10 imbalanced classes, subspace clustering) | clustering accuracy (%): 60.4 -> 70.9 | Improved |
| use tNN graph with spectral clustering instead of k-means on representation vectors (ablation) | ESC-FFS · Extended Yale B (10 imbalanced classes, subspace clustering) | clustering accuracy (%): 41.6 -> 71.1 | Improved |
| select exemplars with FFS instead of random for SRC classification | SRC · Extended Yale B (exemplar classification, 100 exemplars, SRC) | classification accuracy (%): 85.4 -> 92.1 | Improved |
| increase lambda in FFS from 20 to 200 (ablation) | FFS · Extended Yale B (exemplar classification) | classification accuracy (%): 20 -> 200 | Improved |
Closest papers by meaning
- Scalable Exemplar-based Subspace Clustering on Class-Imbalanced Data (ECCV 2018 · similarity 0.90)
- Subspace Clustering through Sub-Clusters (JMLR 2021 · similarity 0.85)
- Automatic Model Selection in Subspace Clustering via Triplet Relationships (AAAI 2018 · similarity 0.85)
- Leveraging Union of Subspace Structure to Improve Constrained Clustering (ICML 2017 · similarity 0.85)
- Affine Subspace Robust Low-Rank Self-Representation: From Matrix to Tensor (TPAMI 2023 · similarity 0.84)
- Convergence and Recovery Guarantees of the K-Subspaces Method for Subspace Clustering (ICML 2022 · 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.
