Edit Question - KnowledgeHub

Title Body

I want to slice a NumPy nxn array. I want to extract an arbitrary selection of m rows and columns of that array (i.e. without any pattern in the numbers of rows/columns), making it a new, mxm array. For this example let us say the array is 4x4 and I want to extract a 2x2 array from it. Here is our array: from numpy import * x = range(16) x = reshape(x,(4,4)) print x [[ 0 1 2 3] [ 4 5 6 7] [ 8 9 10 11] [12 13 14 15]] The line and columns to remove are the same. The easiest case is when I want to extract a 2x2 submatrix that is at the beginning or at the end, i.e. : In [33]: x[0:2,0:2] Out[33]: array([[0, 1], [4, 5]]) In [34]: x[2:,2:] Out[34]: array([[10, 11], [14, 15]]) But what if I need to remove another mixture of rows/columns? What if I need to remove the first and third lines/rows, thus extracting the submatrix [[5,7],[13,15]] ? There can be any composition of rows/lines. I read somewhere that I just need to index my array using arrays/lists of indices for both rows and columns, but that doesn't seem to work: In [35]: x[[1,3],[1,3]] Out[35]: array([ 5, 15]) I found one way, which is: In [61]: x[[1,3]][:,[1,3]] Out[61]: array([[ 5, 7], [13, 15]]) First issue with this is that it is hardly readable, although I can live with that. If someone has a better solution, I'd certainly like to hear it. Other thing is I read on a forum that indexing arrays with arrays forces NumPy to make a copy of the desired array, thus when treating with large arrays this could become a problem. Why is that so / how does this mechanism work?

Tags (comma-separated) Cancel