Numpy multi-dimensional array slicing from end to first

Question

From a numpy array

a=np.arange(100).reshape(10,10)

I want to obtain an array

[[99, 90, 91],
 [9, 0, 1],
 [19, 10, 11]]

I tried

a[[-1,0,1],[-1,0,1]]

but this instead gives array([99, 0, 11]) . How can I solve this problem?

Answer 1

a[[-1,0,1],[-1,0,1]] This is wrong, this means you want an elements from row -1 , column -1 ie (99) and row 0 , column 0 ie (0) and row 1 , column 1 ie (11) this is the reason you are getting array([99, 0, 11])

Your Answer:

a[ [[-1],[0],[1]], [-1,0,1] ] : This means, we want every element from column -1, 0, 1 from row [-1], [0], [1] .

Answer 2

Roll your array over two axis and slice 3x3:

>>> np.roll(a, shift=1, axis=[0,1])[:3, :3]
array([[99, 90, 91],
       [ 9,  0,  1],
       [19, 10, 11]])

Answer 3

Split the slicing into two seperate operations

arr[ [ -1,0,1] ][ :, [ -1,0,1]]
# array([[99., 90., 91.],
#        [ 9.,  0.,  1.],
#        [19., 10., 11.]])

Equivalent to:

temp = arr[ [ -1,0,1] ]  # Extract the rows
temp[ :, [ -1,0,1]]      # Extract the columns from those rows
# array([[99., 90., 91.],
#        [ 9.,  0.,  1.],
#        [19., 10., 11.]])