I am a little confused with pca, there are pca, pca compute, pca apply and pca invertion, I am wondering how to use these nodes? I am going to use pca do cross vallidation
so basically pca is just a projection.
It can, however, be split into two parts:
(i) determining how to project
(ii) execute the projection
the pca node does both at once, while the learner executes only (i) and the pca apply executes (ii).
The advantage is, that you can use a small data set to learn the projection and then apply it to a larger dataset or to new, incoming data.
The pca inversion just reverts the projection.
Note that since the projection is not lossless (depending on your settings) the original data is not reproduced exactly but only approximately according to the information loss you configured in the node settings of the pca-learner.
hope that helps