Research map / Learning theory and algorithms / Matrix and tensor methods
The Search Problem in Mixture Models
Avik Ray, Joe Neeman, Sujay Sanghavi, Sanjay Shakkottai
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 side-information Whitening search instead of seeded k-means | GMM image segmentation · BSDS500 test | median NMI: 0.13 -> 0.17 | Improved |
| use Whitening instead of seeded k-means | GMM image segmentation · BSDS500 train | median NMI: 0.15 -> 0.15 | No clear effect |
| use Whitening instead of seeded k-means | GMM image segmentation · BSDS500 val | median NMI: 0.09 -> 0.11 | No clear effect |
| use Whitening instead of seeded k-means | GMM image segmentation · BSDS500 test | median runtime (s): 81.5 -> 6.7 | Improved |
| use Whitening instead of tensor power method | topic model on NY Times (k=10) · NY Times | runtime (s): 908 -> 188 | Improved |
| use Whitening instead of NMF and SS-NMF | topic modeling with labeled words · NY Times / Yelp | PMI score: 40 of 62 words; 35 of 54 words | Improved |
| use Whitening instead of NMF and SS-NMF | topic modeling · NY Times / Yelp | speedup: more than 2X | Improved |
| use Whitening/Cancellation instead of TPM | GMM synthetic · k=10, d=500 | relative error gain | Improved |
| initialize tensor power iteration with side information (Fast-TPM) (ablation) | TPM · GMM synthetic | error gain over TPM | No clear effect |
| increase sample size n, proportion alpha and variance sigma (ablation) | Whitening · GMM synthetic | error gain over TPM | Improved |
| use Whitening on rare component (alpha_min .0037) | Whitening · GMM k=10, d=500 | error gain over TPM | Improved |
| use subspace search instead of k-means | subspace clustering · Synthetic k=5, m=10, d=500 | speedup: over 4X | Improved |
Closest papers by meaning
- Mixtures Closest To A Given Measure: A Semidefinite Programming Approach (ICML 2026 · similarity 0.87)
- Mix and Match: An Optimistic Tree-Search Approach for Learning Models from Mixture Distributions (NeurIPS 2020 · similarity 0.87)
- Exploration of the Search Space of Gaussian Graphical Models for Paired Data (JMLR 2024 · similarity 0.86)
- Achieving Optimal Clustering in Gaussian Mixture Models with Anisotropic Covariance Structures (NeurIPS 2024 · similarity 0.86)
- Big Learning Expectation Maximization (AAAI 2024 · similarity 0.85)
- Toward Global Convergence of Gradient EM for Over-Paramterized Gaussian Mixture Models (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.
