CODEWITHSUNDEEP
pythonnumpyscipy
monkeyking
monkeyking

Reputation: 6968

extract columns from matrix

How do I extract the first ncolumns from a matrix, or how do i extract the columns from say, column 5 to column 5+n ?

The scenario is that I have a matrix, of dim=(i,j), then I want to look a contiguous blocks of column of length n. So I want to look at the matrix from column0 to column n, then at column1 to column n+1. etc. Until I reach the end of my matrix.

Thanks

>>> mat=array(range(0,20)).reshape(2,10)
>>> mat
array([[ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9],
       [10, 11, 12, 13, 14, 15, 16, 17, 18, 19]])

Given n=8. I want to extract the first 8 columns.

Then the matrix from column [1,11] to column [8,18]

Then the matrix from column [2,12] to column [9,19]

Upvotes: 2

Views: 6465

Answers (2)

unutbu
unutbu

Reputation: 880797

Building on Sven Marnach's answer, if you want to perform a calculation on mat[:,0:8], then on mat[:,1:9], etc, you might be able to perform all the calculations "at once" on the 3-dimensional matrix

In [119]: x = np.lib.stride_tricks.as_strided(mat, shape = (3,2,8), strides = mat.itemsize*np.array([1,mat.shape[-1],1]))

In [120]: x
Out[120]: 
array([[[ 0,  1,  2,  3,  4,  5,  6,  7],
        [10, 11, 12, 13, 14, 15, 16, 17]],

       [[ 1,  2,  3,  4,  5,  6,  7,  8],
        [11, 12, 13, 14, 15, 16, 17, 18]],

       [[ 2,  3,  4,  5,  6,  7,  8,  9],
        [12, 13, 14, 15, 16, 17, 18, 19]]])

For example, to find the inner product of the rows, you could compute all three with 

In [127]: np.sum(x[:,0,:]*x[:,1,:],axis = -1)
Out[127]: array([420, 564, 724])

instead of using a Python loop:

In [128]: for i in range(3):
   .....:     print(np.sum(mat[0,i:i+8]*mat[1,i:i+8]))
   .....:     
   .....:     
420
564
724

Upvotes: 3

Sven Marnach
Sven Marnach

Reputation: 602635

>>> mat[:, 0:8]
array([[ 0,  1,  2,  3,  4,  5,  6,  7],
       [10, 11, 12, 13, 14, 15, 16, 17]])
>>> mat[:, 1:9]
array([[ 1,  2,  3,  4,  5,  6,  7,  8],
       [11, 12, 13, 14, 15, 16, 17, 18]])

Upvotes: 4

Related Questions