CODEWITHSUNDEEP
pythonarraysnumpy
mattiav27
mattiav27

Reputation: 685

matrix product with numpy.dot()

I am trying to calculate the product:

tA*M*B

where A, B are two vectors and M is a squared matrix, tA is the transposed A. The result should be a number.

Numpy has the dot() function that multiplies arrays and matrix: is there a way I can use it to calculate my product in a single blow?

I am using python 2.6

Upvotes: 0

Views: 2253

Answers (3)

hpaulj
hpaulj

Reputation: 231385

np.einsum gives more control over dot operations. There's some debate, though, about when it is faster or slow than np.dot, and whether it consumes too much memory (when the matricies are very big

A=np.arange(1,4)
B=10*np.arange(3,6)
M=np.arange(9).reshape(3,3)
np.dot(A,np.dot(M,B))
np.einsum('i,ij,j',A,M,B)

Upvotes: 1

TooTone
TooTone

Reputation: 8126

You can use the reduce function

reduce(numpy.dot,[tA,M,B])

This is equivalent to

numpy.dot(numpy.dot(tA,M),B)

From the tutorial

reduce(function, sequence) returns a single value constructed by calling the binary function function on the first two items of the sequence, then on the result and the next item, and so on.

An easy to understand example from the documentation is

reduce(lambda x, y: x+y, [1, 2, 3, 4, 5]) calculates ((((1+2)+3)+4)+5)

Whether it's worth using reduce in your case is debatable. However it clarifies the code if you have a long string of matrix multiplications. Compare the following, equivalent lines of code that multiply tA, M1, M2, M3, and B together.

print numpy.dot(numpy.dot(numpy.dot(numpy.dot(tA,M1),M2),M3),B)
print reduce(numpy.dot,[tA,M1, M2, M3,B])

Upvotes: 2

Nitish
Nitish

Reputation: 7126

How about:

import numpy

#Generate Random Data
M = numpy.random.normal(0,1,9).reshape(3,3)
A = numpy.random.normal(0,1,3)
B = numpy.random.normal(0,1,3)

#The Operation
numpy.dot(A, numpy.dot(M,B) )

Upvotes: 2

Related Questions