Fast principal component analysis for high dimensional data implemented according to "Pattern Recognition and Machine Learning" by C. Bishop. For high-dimensional data, fastpca.m is much faster than Python's or Matlabs PCA implementations. Decrease in computation time comes from calculating the PCs from the "small" sample-sample covariance matrix instead of the larger feature-feature covariance matrix. PCs from both covariance matrices can be converted into each other.