Research map / Learning theory and algorithms / Clustering
Efficient Clustering for Stretched Mixtures: Landscape and Optimality
Kaizheng Wang, Yuling Yan, Mateo Diaz
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 |
|---|---|---|---|
| Learn an affine projection that clusters transformed samples around ±1 using uncoupled regression | Fashion-MNIST T-shirt/top vs pullover · Fashion-MNIST training, class ratios 1:1 through 4:1 | misclassification rate: 5.2±0.2%, 7.1±0.4%, 9.3±0.7%, 11.3±1.1% | Improved |
| Use CURE instead of K-means | Fashion-MNIST T-shirt/top vs pullover · Fashion-MNIST training, class ratios 1:1 through 4:1 | misclassification rate: 5.2±0.2% vs 45.1%; 7.1±0.4% vs 49.7%; 9.3±0.7% vs 46.8%; 11.3±1.1% vs 45.1% | Improved |
| Use CURE instead of vanilla spectral clustering | Fashion-MNIST T-shirt/top vs pullover · Fashion-MNIST training, class ratios 1:1 through 4:1 | misclassification rate: 5.2±0.2% vs 42.2%; 7.1±0.4% vs 46.9%; 9.3±0.7% vs 49.7%; 11.3±1.1% vs 49.0% | Improved |
| Use CURE instead of Gaussian-kernel spectral clustering | Fashion-MNIST T-shirt/top vs pullover · Fashion-MNIST training, class ratios 1:1 through 4:1 | misclassification rate: 5.2±0.2% vs 49.9%; 7.1±0.4% vs 33.4%; 9.3±0.7% vs 25.0%; 11.3±1.1% vs 20.0% | Improved |
| Use CURE instead of discriminative K-means | Fashion-MNIST T-shirt/top vs pullover · Fashion-MNIST, N1=1000, class ratios 1:1 through 4:1 | misclassification rate: 5.2±0.3%, 6.7±0.6%, 9.1±0.9%, 11.2±1.2% vs 6.6%, 44.2%, 47.4%, 41.8% | Improved |
| Use CURE instead of the compared clustering method | Fashion-MNIST T-shirt/top vs pullover · Fashion-MNIST, N1=6000, class ratios 1:1 through 4:1 | misclassification rate: 5.2±0.2%, 7.1±0.4%, 9.3±0.7%, 11.3±1.1% vs 31.1±13.8%, 32.9±13.3%, 34.7±12.7%, 36.8±11.2% | Improved |
| Use CURE instead of Mclust | Fashion-MNIST T-shirt/top vs pullover · Fashion-MNIST, N1=6000, class ratios 1:1 through 4:1 | misclassification rate: 5.2±0.2%, 7.1±0.4%, 9.3±0.7%, 11.3±1.1% vs 48.7±1.3%, 39.1±4.8%, 34.1±8.0%, 28.2±7.8% | Improved |
| Use CURE instead of Projection Pursuit | Fashion-MNIST T-shirt/top vs pullover · Fashion-MNIST, N1=6000, class ratios 1:1 through 4:1 | misclassification rate: 5.2±0.2%, 7.1±0.4%, 9.3±0.7%, 11.3±1.1% vs 36.9±9.8%, 37.4±9.6%, 39.7±6.9%, 40.6±7.3% | Improved |
| Use CURE instead of LDA-guided K-means | Fashion-MNIST T-shirt/top vs pullover · Fashion-MNIST, N1=6000, class ratios 1:1 through 4:1 | misclassification rate: 5.2±0.2%, 7.1±0.4%, 9.3±0.7%, 11.3±1.1% vs 45.9%, 49.0%, 45.6%, 44.3% | Improved |
| Use CURE instead of MDH | Fashion-MNIST T-shirt/top vs pullover · Fashion-MNIST, N1=6000, class ratios 1:1 through 4:1 | misclassification rate: 5.2±0.2%, 7.1±0.4%, 9.3±0.7%, 11.3±1.1% vs 48.6%, 43.1%, 38.3%, 35.2% | Improved |
Closest papers by meaning
- Clustering Small Samples With Quality Guarantees: Adaptivity With One2all PPS (AAAI 2018 · similarity 0.88)
- Efficient Clustering Based On A Unified View Of $K$-means And Ratio-cut (NeurIPS 2020 · similarity 0.87)
- A Sampling-Based Approach for Efficient Clustering in Large Datasets (CVPR 2022 · similarity 0.87)
- Achieving Optimal Clustering in Gaussian Mixture Models with Anisotropic Covariance Structures (NeurIPS 2024 · similarity 0.87)
- Robust Discriminative Clustering with Sparse Regularizers (JMLR 2017 · similarity 0.87)
- Adaptive Clustering through Semidefinite Programming (NeurIPS 2017 · similarity 0.87)
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.
