neurodata / non-parametric-clustering

Gui's repo
5 stars 2 forks source link

Implement Procustes or Other Kind of Distance in K-Medoids and K-Means #8

Closed guisf closed 8 years ago

guisf commented 8 years ago

Procrustes distance was used within K-medoids algorithm. The notebook is http://nbviewer.jupyter.org/github/neurodata/non-parametric-clustering/blob/master/procrustes.ipynb

For the MNIST dataset this gives a slightly worst result compared to purely Euclidean metric. This data set is not a "shape". Maybe we should try other dataset.