Python：在numpy数组上按元素明智的操作构建矩阵

Question

Given two numpy arrays (arr1 and arr2) I want to build a matrix that stores the difference of each element in arr1 with each element in arr2. ie:

my_matrix = [arr1-i for i in arr2]

However this starts to get slow as the arrays get larger in size. I've tried to make use of numpy's good performance like so:

arr1_mtx = np.array([arr1]*len(arr2))
arr2_mtx = np.array([arr1]*len(arr2)).T
my_matrix = arr1_mtx-arr2_mtx

I'm quite new to Python so I'm not sure if this is the most pythonic and efficient way to build this matrix. Any tips?

Thanks in advance!

Answer 1

You can reshape one of the arrays to a 2d array, then make use of numpy broadcasting :

arr1[:,None] - arr2

---

arr1 = np.array([1,2])
arr2 = np.array([3,4,5])
​
arr1[:,None] - arr2
#array([[-2, -3, -4],
#       [-1, -2, -3]])