Changed u & sigma to correspond w/ MATLAB code

[benjoch]

When looking at the sqrt(1/sigma) problem I referenced the MATLAB code and noticed differences between the MATLAB and R functions. Namely:

• [U, SIGMA] = eig() in MATLAB will return a matrix of the right eigenvectors to U and a diagonal matrix of the eigenvalues to SIGMA. So our 'u' in the R script should correspond to the eigenvectors and our 'sigma' to the eigenvalues
• diag() in MATLAB will return a column vector of the main diagonal elements of A when given a square matrix so since SIGMA is a diagonal matrix of eigenvalues in MATLAB diag(SIGMA) returns a column vector of the eigenvalues. In the R script, eigendecomp$values returns a vector of the eigenvalues not a diagonal matrix of the eigenvalues so the diag() call is not needed in R

However, after making these fixes I still had the same issue with sqrt(1/sigma) which I am working on. Let me know if I'm making a dumb mistake here. Also here's the MATLAB documentation about the functions that I referenced:
eig() - https://www.mathworks.com/help/matlab/ref/eig.html#d123e352209
diag() - https://www.mathworks.com/help/matlab/ref/diag.html

[thisjustinh]

Very good catch; I'll make these changes in my code and push my stuff now however so this branch won't be merged.

With these changes the values for fx and fy start to directly match what I'm getting in MATLAB but like you say the negative sqrt is still problematic.