Train kmeans in Java API for BOW

asked 2013-07-25 08:47:15 -0500

EvanSimpson gravatar image

Since the BOWTrainer currently is not implemented in the Java API I'm just recreating all of the steps manually, however I'm having trouble with creating a training model that can be reused on independent test data, specifically the clustering of feature vectors.

The kmeans operator will give you the resulting labels and the found centroids, but it does not provide any way of using those centroids to then cluster any new data.

At this point do I just need to implement my own version of kmeans that uses these center points, or is there a method somewhere in the Java API that will do this for me?

edit retag flag offensive close merge delete

Comments

@EvanSimpson Have you managed to find a workaround?

basilevs gravatar imagebasilevs ( 2014-06-14 01:40:53 -0500 )edit