Vastly improve the quality of clusterings.

Setting a minimum cluster size prevents the algorithm from trying to 'cheat' by creating many 3- or 4-colour clusters. This was especially bad when PCA would determine only two basis functions were necessary (the third one being essentially random), causing massive reconstruction errors.

1 files changed, 1 insertions, 1 deletions

diff --git a/demo.py b/demo.py
index 10e6251..dbaa77d 100644
--- a/demo.py
+++ b/demo.py
@@ -17,7 +17,7 @@ if __name__ == '__main__':
 
     print('Clustering...')
     before = clustering.root.total_reconstruction_error()
-    clustering.optimise(8)
+    clustering.optimise()
     after = clustering.root.total_reconstruction_error()
 
     print('Error before: %g' % before)