I know i have asked similar question, but still not sure i have got appropriate reply.I have separate train and test data sets. The feature size is high , 400. I performed PCA on train and test data to reduce feature dimension before classification using SVM. My doubt is do i need to convert the reduced size features with Principle Components as axes to original axes using command such as
OriginalData=RowFeatureVector' * FinalTransformedData + Mean
where RowFeatureVector' isthe matrix of eigenvectors as columns and FinalTransformedData is the transformed data and Mean is the vector of means.
OR i can procced with classification using reduced features in principal component axes?
Help for the same will be highly appreciated as my work is held up.