I always thought that "Eigenman" would make a cool superhero. A man derived from the covariance matrix of the probability distribution of the high-dimensional vector space of all of mankind. His superpowers would be instant recognition of anyone he meets, self-supervised learning, and the ability to do a Karhunen–Loève transform in his head.
Principal component analysis at it's best. Works awfully well on most machine learning problems. Often PCA is not just enough by itself, so you run it before you run a more sophisticated algorithm (say k-Nearest-Neighbors, or Support Vector Machines). It helps in reducing the training time and execution time of those algorithms.
Seeing the word Eigenface brings back chills from my senior year computer vision class. It wasn't that the class was not useful, it was just that the professor was painfully horrible. Although, the eigenface project at the end of the year was VERY interesting. If I can pull up the old code, maybe I'll put it up on github. It was MATLAB code