You had to store tons and tons of tiny values.
Then along came a guy named Dr. Karl Pearson,
Said you can store signals in vectors, just as well.
Just put your data together in a matrix,
Differences and covariances too,
Decompositing, transposing and truncating,
It's amazing what mathematicians with the dot product will do!
PCA, when you need to measure variation (measure variation)
PCA, when you need to waste some time (waste some time)
PCA, when you need to find out who the outlier is (who's the outlier?)
PCA, when you can't afford a spline (afford a spline)
To all the computer scientists out there doing PCA, the singular-value decomposition salutes you, with its all-new line of terrifyingly complex linear algebraic methods for squaring large groups of numbers via the dot product.