PCA is used for dimensional reduction. I learned today that PCA cannot be used for nonlinear data. When nonlinear, you have to use kernel PCA (KPCA).
It seems that since KPCA is more applicable to more variety of data, it seems like a superior method, other than its longer computation time. Is KPCA truly superior to PCA, or are there any practical disadvantages of using KPCA when data is linear?
It seems that if computational efficiency isn't that big of a concern, I can just almost always naively use KPCA.