Mean Shift
Density-based clustering that automatically discovers the number of clusters
Mean Shift iteratively moves each data point toward the densest region in its neighborhood until convergence, forming cluster centers. It does not require the number of clusters to be specified.
When to use:
- When the number of clusters is unknown and automatic discovery is desired
- Smooth, blob-shaped clusters in low-to-medium dimensional spaces
- Relatively small datasets (O(n²) complexity)
Input: Tabular data with the feature columns defined during training Output: Cluster label for each row
Model Settings (set during training, used at inference)
Bandwidth (default: auto-estimated) Kernel bandwidth controlling cluster size. Smaller bandwidth → more, smaller clusters; larger bandwidth → fewer, larger clusters. Auto-estimation works well in most cases.
Bin Seeding (default: false) If true, initializes kernel locations using binned data for faster convergence on large datasets.
Cluster All (default: true) If true, all points (including low-density "orphans") are assigned to the nearest cluster. If false, orphans are labeled -1.
Inference Settings
No dedicated inference-time settings. New points are assigned to the nearest trained cluster center.