|
I was trying the CCA module in PLS and comparing the results with those of CCA package in R and canoncor command in MATLAB. R and MATLAB results agree but the scikit learn results seem to be totally off? Does anyone has any clue why this is happening? I am familiar with the theory of CCA -- is there something fundamentally different that I am missing? |
Maybe you should email this question to the scikit mailing list with an example dataset where things fail?
Yes, I did. Wei Liu from Hong Kong University said the same as you. I have mailed him the dataset but yet to hear back from him :)