Clustering experiments group time series with similar deformation profiles. We contrast a purely temporal baseline against graph-based smoothing strategies that inject spatial consistency.

Baseline

Fast and scalable clustering over dynamic and static features.

• Batched K-means - mini-batch centroid updates for scalable clustering on large datasets.

Graph-Aware Extension

Inject neighborhood information to reduce local noise and improve cluster coherence.

• Spatial-Smoothing Clustering - graph-based smoothing before mini-batch clustering.