Comment by jlrubin

I appreciate that the author wanted to implement their own eigen vector/value method, but really they should use:

    numpy.linalg.eig(x)[0]
    numpy.linalg.eigvals(x)[0]

Numerical stability can be hard to get right...