silvaan
Simulations/Machine Learning/Hierarchical Clustering
Machine Learning

Hierarchical Clustering

Instead of committing to a number of clusters, merge the two closest groups over and over until everything is one. Step through the merges and watch the dendrogram grow branch by branch, or drop your own points and cluster them.

Merges done0
Clusters—

Click the left half to add points. Restart, then Step, to merge the two nearest groups one at a time.

What to observe

  1. Each Step joins the two nearest groups and records how far apart they were as the bar's height. Short bars mean tight, obvious groups; tall bars join things only reluctantly.
  2. You never chose a number of clusters. One tree holds them at every scale: stop early for many small groups, run to the end for one.
  3. Switch the linkage. Single-link chains through bridges and can merge far things via stepping stones; complete-link keeps clusters compact. Same points, different tree.

Shortcuts: space run/pause · s step · r reset · f fullscreen