Arrays can also be split into multiple arrays along the horizontal, vertical, and depth axes using the np.hsplit(), np.vsplit(), and np.dsplit() functions. We will only look at the np.hsplit() function as the others work similarly.
The np.hsplit() function takes the array to split as a parameter, and either a scalar value to specify the number of arrays to be returned, or a list of column indexes to split the array upon.
If splitting into a number of arrays, each array returned will have the same count of columns. The source array must have a number of columns that is a multiple of the specified value.
To demonstrate this, we will use the following array with four columns and three rows:
In [70]: # sample array a = np.arange(12).reshape(3, 4) a Out[70]: array([[ 0, 1, 2, 3], [ 4, 5, 6, 7], [ 8, 9, 10, 11]])
We can split this into four arrays, each representing the values in a specific column:
In [71]: # horiz split the 2-d array into 4 array columns np.hsplit(a, 4) Out[71]: [array([[0], [4], [8]]), array([[1], [5], [9]]), array([[ 2], [ 6], [10]]), array([[ 3], [ 7], [11]])]
Using a value of 2 returns two matrices with two columns each:
In [72]: # horiz split into two array columns np.hsplit(a, 2) Out[72]: [array([[0, 1], [4, 5], [8, 9]]), array([[ 2, 3], [ 6, 7], [10, 11]])]
Also, the following code splits an array along specific columns:
In [73]: # split at columns 1 and 3 np.hsplit(a, [1, 3]) Out[73]: [array([[0], [4], [8]]), array([[ 1, 2], [ 5, 6], [ 9, 10]]), array([[ 3], [ 7], [11]])]
The np.split() function performs an identical task when using axis=1:
In [74]: # along the rows np.split(a, 2, axis = 1) Out[74]: [array([[0, 1], [4, 5], [8, 9]]), array([[ 2, 3], [ 6, 7], [10, 11]])]
Vertical splitting works similarly to horizontal splitting, except against the vertical axis, which can be seen here:
In [75]: # new array for examples a = np.arange(12).reshape(4, 3) a Out[75]: array([[ 0, 1, 2], [ 3, 4, 5], [ 6, 7, 8], [ 9, 10, 11]])
We can split this by 4 and get the four arrays representing the rows:
In [76]: # split into four rows of arrays np.vsplit(a, 4) Out[76]: [array([[0, 1, 2]]), array([[3, 4, 5]]), array([[6, 7, 8]]), array([[ 9, 10, 11]])]
Alternately, splitting by 2, retrieving two arrays of two rows each:
In [77]: # into two rows of arrays np.vsplit(a, 2) Out[77]: [array([[0, 1, 2], [3, 4, 5]]), array([[ 6, 7, 8], [ 9, 10, 11]])]
Splitting can also be performed on specific rows:
In [78]: # split along axis=0 # row 0 of original is row 0 of new array # rows 1 and 2 of original are row 1 np.vsplit(a, [1, 3]) Out[78]: [array([[0, 1, 2]]), array([[3, 4, 5], [6, 7, 8]]), array([[ 9, 10, 11]])]
Likewise, the split command does the same when specifying axis=0:
In [79]: # split can specify axis np.split(a, 2, axis = 0) Out[79]: [array([[0, 1, 2], [3, 4, 5]]), array([[ 6, 7, 8], [ 9, 10, 11]])]
Depth splitting splits three-dimensional arrays. To demonstrate this, we will use the following three-dimensional array:
In [80]: # 3-d array c = np.arange(27).reshape(3, 3, 3) c Out[80]: array([[[ 0, 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]]])
This array can be depth split by 3:
In [81]: # split into 3 np.dsplit(c, 3) Out[81]: [array([[[ 0], [ 3], [ 6]], [[ 9], [12], [15]], [[18], [21], [24]]]), array([[[ 1], [ 4], [ 7]], [[10], [13], [16]], [[19], [22], [25]]]), array([[[ 2], [ 5], [ 8]], [[11], [14], [17]], [[20], [23], [26]]])]