Comment by tel

PCA goes far deeper than meets the eye. For instance, it's a well-known phenomenon that too much dimensionality can actually drive predictor performance to random, but PCA can mitigate that. It's a basically the bread and butter of practical unsupervised learning.