Clustering on MNIST (ARI)
0.96ARIPLSCAN
Evaluation results
| Method | Links | |
|---|---|---|
| PLSCANk=4, Cluster Selection Strategy=Top-5 most persistent layer, Persistence Measure=size, Dimensionality Reduction (UMAP)=min(50, D)2025.12 | 0.96 | |
| IDECPoints=70000, Dim=784, Clusters=10, runs=102026.02 | 0.8 | |
| KBCPoints=70000, Dim=784, Clusters=10, runs=102026.02 | 0.77 | |
| HDBSCAN*k=4, Cluster Selection Strategy=EOM, Persistence Measure=N/A, Dimensionality Reduction (UMAP)=min(50, D)2025.12 | 0.66 | |
| PLSCANk=4, Cluster Selection Strategy=Top-5 most persistent layer, Persistence Measure=size-d, Dimensionality Reduction (UMAP)=min(50, D)2025.12 | 0.66 | |
| PLSCANk=4, Cluster Selection Strategy=Top-5 most persistent layer, Persistence Measure=size-lambda, Dimensionality Reduction (UMAP)=min(50, D)2025.12 | 0.66 | |
| PLSCANk=4, Cluster Selection Strategy=Top-5 most persistent layer, Persistence Measure=lambda, Dimensionality Reduction (UMAP)=min(50, D)2025.12 | 0.66 | |
| FedKmeansKnown K?=Yes2026.01 | 0.64 | |
| CCPoints=70000, Dim=784, Clusters=10, runs=102026.02 | 0.62 | |
| FedGEMKnown K?=No2026.01 | 0.452 | |
| k-meansPoints=70000, Dim=784, Clusters=10, runs=102026.02 | 0.38 | |
| k-FEDKnown K?=Yes2026.01 | 0.354 | |
| FFCM-avg2Known K?=Yes2026.01 | 0.336 | |
| GMM (central)Known K?=Yes2026.01 | 0.287 | |
| FFCM-avg1Known K?=Yes2026.01 | 0.148 | |
| DP-GMM (central)Known K?=No2026.01 | 0.115 | |
| AFCLKnown K?=No2026.01 | 0.038 | |
| HDBSCAN*k=4, Cluster Selection Strategy=leaf, Persistence Measure=N/A, Dimensionality Reduction (UMAP)=min(50, D)2025.12 | 0.01 | |
| PLSCANk=4, Cluster Selection Strategy=Top-5 most persistent layer, Persistence Measure=d, Dimensionality Reduction (UMAP)=min(50, D)2025.12 | 0.01 |