Canonical correlations for test data?

gallantlab / pyrcca

Regularized kernel canonical correlation analysis in Python

Other

242 stars 73 forks source link

Canonical correlations for test data? #8

Open kimrants opened 6 years ago

kimrants commented 6 years ago

After running validate, is it possible to obtain the canonical correlations for the components? As far as I can read, you can only get the correlation to the actual test data and/or predictions... but wouldn't it better for validating the model to obtain the canonical correlations?

Thanks!

kimrants commented 6 years ago

I actually don't understand why the validation method that adds the attribute: cca.corrs is supposed to work? As far as I understand, CCA produces weights w_a and w_b such that corr(Aw_a, Bw_b) is maximised.

If you just find the correlation corr(A_test*w_a, B_test) you are not using one of the mappings? I don't see why this should work?

Thank you!

sdrakulich commented 6 years ago

Any followup on this?