<<<
>>>
Principal Component Analysis
Note: we can rotate our coordinate system by some θ to be orthogonal to our data.
Only variance terms will then be non-zero in our 2-D Gaussian tensor
Covariance terms go to zero
