[Numpy-discussion] Tensor-Like Outer Product Question

[matt_in_nyc]

I am trying to perform the following operation:

X is an m by n matrix, and I want to store outer products of the form Y[i] =
numpy.outer(X[i,:], X[i,:]), leading to the relation Y[i,j,k] =
X[i,j]*X[i,k] for i = 0,...,m-1; j,k = 0,...,n-1.  I am trying to think of
how to do this using tensordot, but so far I am finding no inspiration.

Some far, my only solution has been to loop over i

Y = numpy.empty([m,n,n])
for i in range(m):
    Y[i] = numpy.outer(X[i,:], X[i,:])

but this is fairly slow as for my dataset, m is of order 10^7 or 10^8 and n
is around 20.  Any help on how to vectorize/tensorize this operation to
avoid the for loop would be much appreciated.
-- 
View this message in context: http://old.nabble.com/Tensor-Like-Outer-Product-Question-tp29339226p29339226.html
Sent from the Numpy-discussion mailing list archive at Nabble.com.