Repository navigation
Invalid output shape from np.dot with matrixes (Trac #2057) #2649
Description
Original ticket http://projects.scipy.org/numpy/ticket/2057 on 2012-02-17 by @pv, assigned to unknown.
http://permalink.gmane.org/gmane.comp.python.scientific.user/31095
Consider this:
import numpy as np
x = np.arange(5)
I = np.asmatrix(np.identity(5))
print np.dot(I, x).shape
# -> (1, 5)
while (5, 1) would be what is expected based on usual linear algebra rules.
It might also be a mistake to return a matrix from mixed matrix--array computations.
Activity
BUG: solve issue numpy#2649.
charris commentedon Feb 21, 2014
I'm going to close this. This is a topic of endless discussion whenever 'fixing' matrices comes up. The current behaviour is that a 1D vector on the left acts like a row vector, and on the right acts like a column vector. By default, matrix([1,2,3]) is a row vector. Unfortunate, but true, and likely to stay that way until someone rethinks the matrix package.
charris commentedon Feb 21, 2014
If you need row and column vectors, you need to construct them as matrices. Maybe adding functions rowvec, colvec to construct the relevant matrices would help.