Expectation Maximization inital distribution
Hello I do have a questions about the EM inital distribution:
What does initial distribution "average parameters" in the EM-Clustering operator use? The average over all attributes? Sadly it is not specified in the documentation.
Thanks for all future help 
#Edit: Arg, sorry I had clicked to fast for my browser so this post is definitly in the wrong place, can someone who has the power move the question to the studio forum pls?
Answers
-
Good question, I don't know the answer of the top of my head.
Are you good at reading Java? I'm not but the source code is here (line 338): https://github.com/rapidminer/rapidminer-studio/blob/85d3bee36c026a70580075092ed85ac517369e8e/src/main/java/com/rapidminer/operator/clustering/clusterer/soft/EMClusterer.java
0 -
Okay thx, I think I got it.
The method first searches the average, minimum and maximum over all attributes. Then it places half of the initial distributions between the minimum and average, and the other half between average and maximum. All Clusters then get the same std. deviation over all the data points.0