This is more of a research project then some testable project. It's based on the Mika et al paper "Kernel PCA and De-Noising in Feature Space".The code of mine is attached. The implementation via linear PCA in lpca.py and other one in kpca.py One could see that the performance of kernal PCA is much better then that shown by linear PCA. The kernal used in kPCA is gaussian only.