K-Means initialization

KNIME Analytics Platform
1

Hi

How can i change the initialization for the k-means algorithm. The k-means initialization should be random, but aparently it is fixed in the KNIME implementation. How this initialization is done? if it is random with a fixed seed, is it possible to change the seed or set it to be random?

Thansk for help

 

Alexandre Evsukoff

 

 

 

 

 

 

 

 

 

 

2

Hm...

In the code I saw the centres are initialized with the first row in the data table.


You can install the Weka plugin. There is a k-means algorithm as well. I think this one sets random initial centre values.

An other option would be to change the implementation.