CODEWITHSUNDEEP
python-3.xnumpy
detly
detly

Reputation: 30332

How can I do element-wise arithmetic on Numpy matrices?

I am using Numpy's matlib style matrices for a particular algorithm. This means that the multiplication operator * performs the equivalent of an ndarray's dot():

>>> import numpy.matlib as nm
>>> a = nm.asmatrix([[1,1,1],[1,1,1],[1,1,1]])
>>> b = nm.asmatrix([[1,0,0],[0,1,0],[0,0,1]])
>>> a * b
matrix([[1, 1, 1],
        [1, 1, 1],
        [1, 1, 1]])

Is there a method to perform element-wise arithmetic, like the * operator does on ndarrays?

Upvotes: 1

Views: 163

Answers (1)

DSM
DSM

Reputation: 353079

You could use np.multiply:

>>> a = np.matrix(np.random.rand(3,3))
>>> b = np.matrix(np.random.rand(3,3))
>>> a * b
matrix([[ 1.29029129,  0.53126365,  2.12109815],
        [ 0.99370991,  0.55737572,  1.9167072 ],
        [ 0.76268194,  0.43509462,  1.48640178]])
>>> np.asarray(a) * np.asarray(b)
array([[ 0.67445535,  0.12609799,  0.7051103 ],
       [ 0.00131878,  0.42079486,  0.5223201 ],
       [ 0.65558303,  0.03020335,  0.16753354]])
>>> np.multiply(a, b)
matrix([[ 0.67445535,  0.12609799,  0.7051103 ],
        [ 0.00131878,  0.42079486,  0.5223201 ],
        [ 0.65558303,  0.03020335,  0.16753354]])

It's a little unusual to want to perform elementwise multiplication on the same object you're performing matrix multiplication on, but you probably already know that. :^) It might be worthwhile seeing if your algorithm has a nice np.einsum description, though.

Upvotes: 1

Related Questions