problem with result of ICA ?

norah

hello,

i am using ICA in rapidminer but there is a problem when i make the number of components = 6 it is ok but when increase it the data appear in a cells as a question mark and i do not know why? 
even though the number of attributes = 14
is there any standard to decide the number of component ?

varunm1

Hello @norah

I tried with my dataset and observed that we need to normalize rows. In the ICA operator, we can see that there is an option called 'row norm' which needs to be selected. Try this and you will get the reduced dimensions with 6 components.

Thanks 

varunm1

You can use the default values provided by RM. If you want to understand in depth about ICA. Please go through the below paper.

"Independent Component Analysis: Algorithms and Applications"

norah

thanks. ok can you or any one help me with other parameters which are:
function ,
alpha,
max iteration and 
tolerance 
what is mean, how can i decide about it?