Search code examples
pythonarraysnumpydot-product

Dot Product of Two 2-D Column Vectors Using Numpy (Without Transpose)

I am trying to calculate dot product of two 2D vectors using numpy Here is what I know so far:

  1. If we have two 1D np arrays like v and w, we can use np.inner(v, w) to get the dot product of these two vectors but since 1D arrays have shape of (m, ) I don't like to use them due to making troubles by that lost second dimension in (m, )
  2. If we use 2D arrays to define v and w like below:
    v = np.array([[1], [3], [5]])
    w = np.array([[2], [4], [6]])

We can calculate the dot product by transposing one of these vectors and then using matrix multiplication:

    vdotw = v.T @ w

or:

   vdotw = w.T @ v

My question is that is there any method is numpy that specifically calculate dot product of two vectors (defined as 2D arrays not those funny 1D arrays) directly and without transposing one of them?

Solution

  • Numpy's einsum() can be used to do this.

    However, your request is very unusual. Why don't you want to use transpose? The transpose operator has virtually zero cost: it does not make a copy of the underlying data as I suspect you imagine it to.