It looks like your code is wrong, and numpy 1.12 happened to let you get away with it

This line:

evals = evals[evals > tolerance]

Reduces the eigenvalues to only those which are greater than the tolerance

When you do U[:, evals > tolerance], evals > tolerance is just going to be [True, True, ...].

You need to swap the last two lines, to

U = U[:, evals > tolerance]
evals = evals[evals > tolerance]

Or better yet, introduce an intermediate variable:

keep = evals > tolerance
evals = evals[keep]
U = U[:, keep]


On Tue, 12 Feb 2019 at 15:16 Mauro Cavalcanti <> wrote:
Dear ALL,

I am trying to port an eigenalysis function that runs smoothly on Numpy 1.12 but fail miserably on Numpy 1.13 or higher with the dreadful error "boolean index did not match indexed array along dimension 1".

Here is a fragment of the code, where the error occurrs:

    evals, evecs = np.linalg.eig(Syy)
    idx = evals.argsort()[::-1]
    evals = np.real(evals[idx])
    U = np.real(evecs[:, idx])
    evals = evals[evals > tolerance]
    U = U[:, evals > tolerance] # Here is where the error occurs

So, I ask: is there a way out of this?

Thanks in advance for any assistance you can provide.
NumPy-Discussion mailing list