Clustering

The Clustering panel allows to test, tweak and observe how different algorithms perform clustering on samples living in a N-dimensional space: "the canvas".
Clustering can assign a cluster value to each sample, or a contribution of each cluster, or a non-cluster value (for outliers), depending on the algorithm used.

The canvas will display the results of the clustering in multiple layers, which can be toggled using the display options. These are:
In the standard case, a different color is assigned to each cluster. For algorithms providing a contribution from the cluster, a mixing of colors means multiple contributions from different clusters. Be advised that these colors do not correspond to class labels (indeed the data could have no class information whatsoever) but rather they indicate the cluster(s) they have been assigned to.

In Practice
The easiest way to perform clustering is to:
  1. Draw some samples (left-click)
  2. Click on "Cluster"
This should train the algorithm and start painting the canvas with the results of the clustering

Options and Commands
The interface for clustering (the right-hand side of the Algorithm Options dialog) provides the following commands: and the following options: All other options are algorithm-dependent and should be described in the help menu of the algorithm itself.