numerics.eig

[Tim Lupulus]

I already wrote 2 comments on the blog but figured this was a better way to reach you: First of all, great library! I’m afraid I found a bug though, these lines (and probably more) cause internal errors (reproducible in the workshop):

numeric.eig([[1,1,1,1],[0,0,0,0],[0,0,0,0],[0,0,0,0]])

numeric.eig([[1,0,1,0],[0,0,0,0],[0,1,0,1],[0,0,0,0]])
numeric.eig([[1,1,0,1],[0,0,1,1],[0,0,0,0],[0,0,0,0]])

Or am I doing something wrong? Also, it would be great if numerics.eig() could sort the eigenvalues (and corresponding vectors) by their value in a descending manner, like WolframAlpha does. Compare the following line with http://www.wolframalpha.com/input/?i=%5B%5B0.5%2C0.5%2C0%2C0.25%5D%2C%5B0%2C0%2C0.5%2C0.5%5D%2C%5B0.5%2C0.5%2C0%2C0%5D%2C%5B0%2C0%2C0.5%2C0.5%5D%5D -- the results are sorted on WolframAlpha, which is convenient for some cases.

numeric.eig([[0.5,0.5,0,0.25],[0,0,0.5,0.5],[0.5,0.5,0,0],[0,0,0.5,0.5]])

Keep up the good work!