CODEWITHSUNDEEP
pythonarraysnumpyindexingslice
Hamza Ebrahim
Hamza Ebrahim

Reputation: 27

Slicing/Indexing with multidimensional arrays using Numpy

I have 3 4x4 arrays (matrices) created with: arr=np.linspace(1,48,48).reshape(3,4,4)

The matrices appear as below: `

[[[ 1.  2.  3.  4.]
  [ 5.  6.  7.  8.]
  [ 9. 10. 11. 12.]
  [13. 14. 15. 16.]]

 [[17. 18. 19. 20.]
  [21. 22. 23. 24.]
  [25. 26. 27. 28.]
  [29. 30. 31. 32.]]

 [[33. 34. 35. 36.]
  [37. 38. 39. 40.]
  [41. 42. 43. 44.]
  [45. 46. 47. 48.]]]`

I would like to perform indexing/splicing to obtain certain outputs eg: 

[[36. 35.] [40. 39.] [44. 43.] [48. 47.]]

[[13. 9. 5. 1.] [29. 25. 21. 17.] [45. 41. 37. 33.]]


[[25. 26. 27. 28.], [29. 30. 31. 32.], [33. 34. 35. 36.], [37. 38. 39. 40.]]

4*. [[1. 4.] [45. 48.]]

I am struggling with exactly how to approach it. When working with a particular matrix, I have attempted to access that matrix and then splice/index from there. For example, the output [[36. 35.] [40. 39.] [44. 43.] [48. 47.]] lies in the third matrix. I access the matrix like this matrix3 = arr[array([2])]

Now I am only working with rows and columns in the third matrix and am finding it difficult to slice correctly. Should matrix3[::-1,::-1] invert both columns and rows? If yes then is this the best way to approach it? Instead, should I use reshape and should you use reshape on all 3 4x4 arrays or access the matrix you want to work with and then reshape?

edit: added 4.

Upvotes: 0

Views: 215

Answers (2)

hpaulj
hpaulj

Reputation: 231665

Extracting your first result, step by step:

In [53]: arr[2,:,:]            # the desired plane
Out[53]: 
array([[33, 34, 35, 36],
       [37, 38, 39, 40],
       [41, 42, 43, 44],
       [45, 46, 47, 48]])
In [54]: arr[2,:,2:]          # the desired columns
Out[54]: 
array([[35, 36],
       [39, 40],
       [43, 44],
       [47, 48]])
In [55]: arr[2,:,:1:-1]        # the reversed order
Out[55]: 
array([[36, 35],
       [40, 39],
       [44, 43],
       [48, 47]])

or if it's easier, reverse as a separate step:

In [56]: arr[2,:,2:][:,::-1]
Out[56]: 
array([[36, 35],
       [40, 39],
       [44, 43],
       [48, 47]])

2nd

In [57]: arr[:,:,0]          # select column
Out[57]: 
array([[ 1,  5,  9, 13],
       [17, 21, 25, 29],
       [33, 37, 41, 45]])
In [58]: arr[:,::-1,0]        # reverse
Out[58]: 
array([[13,  9,  5,  1],
       [29, 25, 21, 17],
       [45, 41, 37, 33]])
In [59]: arr[:,::-1,0].T      # transpose
Out[59]: 
array([[13, 29, 45],
       [ 9, 25, 41],
       [ 5, 21, 37],
       [ 1, 17, 33]])

3rd

This is a little trickier. We want last 2 rows for one plane, and first two for another plane. To get that we need a pair on indices that will broadcast to the correct shape, pairing up 1 with [2,3] etc.

In [61]: arr[[[1],[2]],[[2,3],[0,1]],:]
Out[61]: 
array([[[25, 26, 27, 28],
        [29, 30, 31, 32]],

       [[33, 34, 35, 36],
        [37, 38, 39, 40]]])

This is a 3d matrix; one way to reduce it to 2d is to concatenate:

In [63]: np.concatenate(arr[[[1],[2]],[[2,3],[0,1]],:],axis=0)
Out[63]: 
array([[25, 26, 27, 28],
       [29, 30, 31, 32],
       [33, 34, 35, 36],
       [37, 38, 39, 40]])

reshape works just as well:

In [65]: arr[[[1],[2]],[[2,3],[0,1]],:].reshape(4,4)
Out[65]: 
array([[25, 26, 27, 28],
       [29, 30, 31, 32],
       [33, 34, 35, 36],
       [37, 38, 39, 40]])

Also arr.reshape(3,2,2,4)[[1,2],[1,0],:].reshape(4,4)

You can write these indexing expressions as tuples with slices, e.g.:

In [66]: idx = (2, slice(None), slice(None,1,-1))
In [67]: arr[idx]
Out[67]: 
array([[36, 35],
       [40, 39],
       [44, 43],
       [48, 47]])

So in general the common tools include indexing (with slices, scalars and lists), reversing (-1 step), transposing (or swapaxes), and reshaping. You can't do everything with just one of these.

Upvotes: 2

twolffpiggott
twolffpiggott

Reputation: 1103

You're on the right track with the slicing! For your desired outputs try:

arr[2,:,:1:-1]
np.vstack((arr[i, ::-1, 0] for i in range(3)))
np.vstack((arr[1, 2:, :], arr[2, :2, :]))

Output:

array([[36., 35.],
   [40., 39.],
   [44., 43.],
   [48., 47.]])

array([[13.,  9.,  5.,  1.],
   [29., 25., 21., 17.],
   [45., 41., 37., 33.]])

array([[25., 26., 27., 28.],
   [29., 30., 31., 32.],
   [33., 34., 35., 36.],
   [37., 38., 39., 40.]])

Upvotes: 0

Related Questions