Research map / Label-efficient and robust learning / Recognition and long-tail
Online Deep Clustering for Unsupervised Representation Learning
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace alternating global clustering and network updates with online batch-wise label reassignment and centroid updates | DeepCluster-style unsupervised representation learning · ImageNet val, AlexNet | top-1 accuracy: 6.7% | Helped |
| Replace alternating global clustering and network updates with online batch-wise label reassignment and centroid updates | DeepCluster-style unsupervised representation learning · VOC2007 SVM | mAP: 9.1% | Helped |
| Replace alternating global clustering and network updates with online batch-wise label reassignment and centroid updates | DeepCluster-style unsupervised representation learning · VOC2007 low-shot SVM | mAP: 10.2% | Helped |
| Fine-tune Colorization with ODC instead of evaluating the original representation | Colorization, ResNet-50 · VOC2007 SVM | mAP: 55.6 -> 72.3; 16.7% | Helped |
| Fine-tune Jigsaw with ODC instead of evaluating the original representation | Jigsaw, ResNet-50 · VOC2007 SVM | mAP: 64.5 -> 74.4; 9.9% | Helped |
| Fine-tune Rotation with ODC instead of evaluating the original representation | Rotation, ResNet-50 · VOC2007 SVM | mAP: 67.4 -> 74.5; 7.1% | Helped |
| Fine-tune DeepCluster with ODC instead of evaluating the original representation | DeepCluster, ResNet-50 · VOC2007 SVM | mAP: 69.1 -> 77.0; 7.9% | Helped |
| Use loss re-weighting with weights proportional to 1/sqrt(class size) instead of uniform sampling | DeepCluster · ImageNet | performance | No clear effect |
| Update the centroids memory less frequently within a reasonable range (ablation) | ODC · VOC2007 SVM | mAP | No clear effect |
| Raise the minimum small-cluster size threshold to about 160 (ablation) | ODC · VOC2007 SVM | mAP | Hurt |
| Vary the minimum small-cluster size within a reasonable range (ablation) | ODC · VOC2007 SVM | mAP | No clear effect |
Closest papers by meaning
- Joint Unsupervised Learning of Deep Representations and Image Clusters (CVPR 2016 · similarity 0.90)
- Hard Regularization to Prevent Deep Online Clustering Collapse without Data Augmentation (AAAI 2024 · similarity 0.88)
- Learning from Sample Stability for Deep Clustering (ICML 2025 · similarity 0.87)
- You Can Trust Your Clustering Model: A Parameter-free Self-Boosting Plug-in for Deep Clustering (NeurIPS 2025 · similarity 0.86)
- P$^2$OT: Progressive Partial Optimal Transport for Deep Imbalanced Clustering (ICLR 2024 · similarity 0.85)
- Exploring a Principled Framework for Deep Subspace Clustering (ICLR 2025 · 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.