Research map / Learning theory and algorithms / Clustering
Dissimilarity-Based Sparse Subset Selection
Ehsan Elhamifar, Guillermo Sapiro, Shankar Sastry
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 |
|---|---|---|---|
| use DS3 instead of AP to select representatives (eta=0.10) | NN classification · 15 Scene Categories (chi2) | error gap (%): 7.91 -> 5.69 | Improved |
| use DS3 instead of AP at the smallest budget (eta=0.05) | NN classification · 15 Scene Categories (chi2) | error gap (%): 11.59 -> 12.04 | Hurt |
| use DS3 instead of Kmedoids (eta=0.35) | NN classification · 15 Scene Categories (chi2) | error gap (%): 7.51 -> 2.90 | Improved |
| use chi2 dissimilarity instead of Euclidean distance (eta=0.35, DS3) (ablation) | DS3 · 15 Scene Categories | error gap (%): 3.24 -> 2.90 | Mixed |
| increase the selected fraction eta from 0.05 to 0.35 (ablation) | DS3 · 15 Scene Categories (chi2) | error gap (%): 12.04 -> 2.90 | Improved |
| initialize SNC with DS3 solution instead of random (eta=0.35) (ablation) | SNC · 15 Scene Categories (chi2) | error gap (%): 1.62 -> -2.53 | Improved |
| initialize SNC with DS3 solution instead of random (eta=0.05) (ablation) | SNC · 15 Scene Categories (chi2) | error gap (%): 10.01 -> 8.52 | Improved |
| select representatives (eta=0.35) rather than use all training samples (ablation) | NN classification, class store · 15 Scene Categories | class accuracy (%): 55.6 -> 61.9 | Improved |
| use DS3 instead of Kmedoids on sequence 2 | motion segmentation · CMU MoCap | clustering error (%): 46.26 -> 9.90 | Improved |
| use DS3 instead of spectral clustering on sequence 9 | motion segmentation · CMU MoCap | clustering error (%): 9.02 -> 11.18 | Hurt |
| vary snippet length Delta for local models (ablation) | DS3 · CMU MoCap | segmentation error | No clear effect |
| use ADMM instead of CVX solver (N=500, p=2) | DS3 optimization · random N x N datasets | computational time (sec): 5.4 x 10^3 -> 3.6 x 10^0 | Improved |
Closest papers by meaning
- Integrative Generalized Convex Clustering Optimization and Feature Selection for Mixed Multi-View Data (JMLR 2021 · similarity 0.86)
- Robust Discriminative Clustering with Sparse Regularizers (JMLR 2017 · similarity 0.85)
- Learning Sparse Combinatorial Representations via Two-stage Submodular Maximization (ICML 2016 · similarity 0.85)
- Subspace Clustering With Priors via Sparse Quadratically Constrained Quadratic Programming (CVPR 2016 · similarity 0.85)
- Approximate Submodularity and its Applications: Subset Selection, Sparse Approximation and Dictionary Selection (JMLR 2018 · similarity 0.85)
- GIST: Greedy Independent Set Thresholding for Max-Min Diversification with Submodular Utility (NeurIPS 2025 · 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.
