Research map / Learning theory and algorithms / Clustering
A Random Matrix Analysis of Data Stream Clustering: Coping With Limited Memory Resources
Hugo Lebeau, Romain Couillet, Florent Chatelain
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 circulant-banded online kernel spectral clustering with memory L=100 instead of offline T×T kernel spectral clustering | online kernel spectral clustering on image streams · BigGAN and Fashion-MNIST image clustering | classification error | Mixed |
| Online method vs batch kernel spectral clustering with L points in memory (same memory) | online kernel spectral clustering · theoretical phase transition (n/p=100) | phase transition position ||mu||^2 | Improved |
| Punctured (offline, Bernoulli mask) kernel spectral clustering compared to online circulant mask at same sparsity | kernel spectral clustering · theoretical phase transition (n/p=100) | phase transition position ||mu||^2 | Improved |
| Use only top eigenvector for binary classification vs additional spike eigenvectors (ablation) | online kernel spectral clustering (binary) · binary Gaussian mixture | classification information | No clear effect |
| Replace k-means on spectral embeddings with sign-based classification of the dominant eigenvector | online kernel spectral clustering · BigGAN and Fashion-MNIST | clustering performance | Improved |
| Increase memory size (ε = (2L-1)/n) for online spectral clustering | online kernel spectral clustering · theoretical phase transition (n/p=100) | phase transition position ||mu||^2 | Improved |
| Average classification over majority vote of n classifications per point vs single-time classification | online kernel spectral clustering · BigGAN and Fashion-MNIST | classification error | No clear effect |
Closest papers by meaning
- Two-way kernel matrix puncturing: towards resource-efficient PCA and spectral clustering (ICML 2021 · similarity 0.90)
- Scalable Multiple Kernel Clustering: Learning Clustering Structure from Expectation (ICML 2024 · similarity 0.90)
- Graph-based Clustering Revisited: A Relaxation of Kernel k-Means Perspective (JMLR 2026 · similarity 0.90)
- Scalable Sequential Spectral Clustering (AAAI 2016 · similarity 0.90)
- Robust Spectral Clustering with Rank Statistics (JMLR 2024 · similarity 0.90)
- Pass-Efficient Algorithms for Graph Spectral Clustering (Student Abstract) (AAAI 2024 · similarity 0.88)
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.
