Research map / Learning theory and algorithms / Clustering
DeepDPM: Deep Clustering With an Unknown Number of Clusters
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 |
|---|---|---|---|
| Infer the cluster count and use split/merge adaptation rather than fixing K | Deep clustering on ImageNet-50 · ImageNet-50 balanced | ACC: .57±.01 -> .64±.00 | Helped |
| Use DeepDPM instead of the strongest reported classical baseline | Clustering with fixed extracted features · MNIST | ACC: .98±.00 -> .98±.00 | No clear effect |
| Use DeepDPM instead of the strongest reported classical baseline | Clustering with fixed extracted features · USPS | ACC: .90±.04 -> .89±.2 | Hurt |
| Use DeepDPM instead of the strongest reported classical baseline | Clustering with fixed extracted features · Fashion-MNIST | ACC: .60±.04 -> .62±.03 | Helped |
| Add AE alternation between clustering and feature learning | DeepDPM on ImageNet-50 · ImageNet-50 balanced | ACC: .64±.00 -> .66±.01 | Helped |
| Add AE alternation between clustering and feature learning | DeepDPM on ImageNet-50 · ImageNet-50 imbalanced | ACC: .58±.01 -> .60±.01 | Helped |
| Disable both splits and merges (ablation) | DeepDPM on Fashion-MNIST · Fashion-MNIST | ACC: .61±.00 -> .59±.03 | Hurt |
| Add splits (ablation) | DeepDPM on Fashion-MNIST · Fashion-MNIST | ACC: .59±.02 -> .59±.00 | No clear effect |
| Add merges (ablation) | DeepDPM on Fashion-MNIST · Fashion-MNIST | ACC: .58±.01 -> .61±.00 | Helped |
| Use learned subclustering nets instead of 2-means (ablation) | DeepDPM on Fashion-MNIST · Fashion-MNIST | ACC: .59±.02 -> .61±.00 | Helped |
| Add priors when computing cluster parameters in the M step (ablation) | DeepDPM on Fashion-MNIST · Fashion-MNIST | ACC: .57±.02 -> .61±.00 | Helped |
| Use Lcl instead of an isotropic loss (ablation) | DeepDPM on Fashion-MNIST · Fashion-MNIST | ACC: .58±.00 -> .61±.00 | Helped |
Closest papers by meaning
- You Can Trust Your Clustering Model: A Parameter-free Self-Boosting Plug-in for Deep Clustering (NeurIPS 2025 · similarity 0.88)
- From Logits to Hierarchies: Hierarchical Clustering made Simple (ICML 2025 · similarity 0.88)
- Hard Regularization to Prevent Deep Online Clustering Collapse without Data Augmentation (AAAI 2024 · similarity 0.87)
- P$^2$OT: Progressive Partial Optimal Transport for Deep Imbalanced Clustering (ICLR 2024 · similarity 0.87)
- Interactive Deep Clustering via Value Mining (NeurIPS 2024 · similarity 0.86)
- Towards Calibrated Deep Clustering Network (ICLR 2025 · similarity 0.86)
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.