milosgajdos/gosom

Cluster labeling

lapnd opened this issue · 2 comments

lapnd commented

Hello,
Thank for your project!
I'm trying to experiment with gosom and I have two questions

  1. How to we labeling cluster after training?
  2. After training, for a new given input vector, how do we know which cluster it belongs to?
    Thank you!
milosgajdos commented

Hi,

  1. This is a rather broad topic, the most useful direction is to generate a heatmap based on how many vectors from the training set fall into particular BMU (Best Matching Unit). Based on the counts of particular BMUs you can generate a heatmap using a code like this
  2. There is a function BMUs which when run against dataset returns a slice of indices of Best Match Unit (BMU) codebook vectors for each vector stored in the data matrix you passed to it
lapnd commented

Thank for your detail code and answer!