[NumPy-Tickets] [NumPy] #1454: numpy.linalg.eig computes wrong eigenvalues

[NumPy Trac]

#1454: numpy.linalg.eig computes wrong eigenvalues
----------------------+-----------------------------------------------------
 Reporter:  skavhaug  |       Owner:  somebody
     Type:  defect    |      Status:  new     
 Priority:  normal    |   Milestone:          
Component:  Other     |     Version:          
 Keywords:            |  
----------------------+-----------------------------------------------------
 When using numpy.linalg.eig to compute the eigenvector/eigenvalue pairs of
 a non-symmetric (and non-Hessian) matrix, the errors defined as:
 np.linalg.norm(np.dot(a, v[:,i]) - w[i]*v[:,i])
 is off by several orders of magnitude (up to 14 orders) compared to when
 performing the same calculation in octave.

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/1454>
NumPy <http://projects.scipy.org/numpy>
My example project