不带for循环的numpy数组求和

Question

I have a 3D numpy(say A) array and want to make this operation without a for loop:

B[0,:,:]=Sum(A,axis=0)-(A[0,:,:])
B[1,:,:]=Sum(A,axis=0)-(A[0,:,:]+A[1,:,:])
B[2,:,:]=Sum(A,axis=0)-(A[0,:,:]+A[1,:,:]+A[2,:,:])

....... and so on... finally B should be a 3D array, each frame of it should be calculated as above.

So I want to calculate B without a for loop.

Sum(A,axis=0) is easy to calculate but the problem in implementing the second term of B without for loop and also append the result to the B(3D array).

could you please help me

Answer 1

You can use np.cumsum :

# To make it easier to write for me, I am going to call you matrix "A": a, and "B": b

a_sum = np.sum(a, axis=0)
a_sum = a_sum[np.newaxis]
a_sum = np.repeat(a_sum, len(b), axis=0)

b = a_sum

a_cumsum = np.cumsum(a, axis=0)
b -= a_cumsum