I can't say why, but for some data sets I get an error while using "PCA Apply" node:
ERROR PCA Apply Execute failed: Matrix inner dimensions must agree.
I can't find any documentation for that. Meanwhile the regular PCA node works without problems...
Could it be a bug or the data set is wrong?
I can't guess what the problem might be from that error message, would it be possible to see an example?
I've preapred a reduced dataset, where the problem still exists. I'm also attaching the (simple) workflow screenshot where this error occurs.
Thanks a lot for the simple example, it made reprodcuing the problem easy.
Just a hint that may lead to a temporary workaround...If you restrict your analysis to the first 3 parameters, the problem is still apparent. If you subsequently remove param1, which is all -1s, things start working as expected. It seems that for some reason this column (and others in the file) are causing the problem.
I'll post back here as soon as I know more.
Thank you very much!
EDIT: As a temporary workaround I've applied low variance filter to eliminate all columns with constant (-1) values and it now works fine.
EDIT2: Above workaround doesn't always work...
I'm attaching another, reduced data set which also causes the problem. There is no longer "-1" columns there.
Maybe it could help you to solve the problem,