NumPy: Setting subset of numpy array at non-NaN positions

Question

I have an array A (m X n) and another W (m X n'), where n' <= n. W can contain nan entries.

I want to set the first n' columns of A with W. However, at the positions where W has a nan, I want original A entries to be preserved.

Here is a sample code.

In [1]: import numpy as np

In [2]: A = np.random.rand(10, 4)

In [3]: A
Out[3]: 
array([[ 0.60879191,  0.13556633,  0.26043647,  0.69874723],
       [ 0.23775325,  0.5744113 ,  0.76446265,  0.82177711],
       [ 0.44320766,  0.43535148,  0.33745034,  0.63270876],
       [ 0.81903997,  0.15170996,  0.9847644 ,  0.77856538],
       [ 0.89771006,  0.11521106,  0.92180393,  0.82296974],
       [ 0.57292616,  0.47236245,  0.643821  ,  0.39011254],
       [ 0.1478904 ,  0.1597816 ,  0.61934679,  0.87229069],
       [ 0.6395053 ,  0.40529633,  0.02389057,  0.0144438 ],
       [ 0.25381449,  0.28045816,  0.98475699,  0.03257699],
       [ 0.91827695,  0.85382925,  0.94231584,  0.5303724 ]])

In [4]: W = np.random.rand(10, 2)

In [5]: W
Out[5]: 
array([[ 0.85731947,  0.02603337],
       [ 0.46941828,  0.12485814],
       [ 0.607665  ,  0.61742206],
       [ 0.67579577,  0.44169912],
       [ 0.77938579,  0.04609614],
       [ 0.55431149,  0.12936694],
       [ 0.05191665,  0.94768561],
       [ 0.9494111 ,  0.21739947],
       [ 0.77785379,  0.35316779],
       [ 0.72959474,  0.72603156]])

In [6]: W[2, 1] = np.NAN

In [7]: W[4, 0] = np.NAN

In [8]: W
Out[8]: 
array([[ 0.85731947,  0.02603337],
       [ 0.46941828,  0.12485814],
       [ 0.607665  ,         nan],
       [ 0.67579577,  0.44169912],
       [        nan,  0.04609614],
       [ 0.55431149,  0.12936694],
       [ 0.05191665,  0.94768561],
       [ 0.9494111 ,  0.21739947],
       [ 0.77785379,  0.35316779],
       [ 0.72959474,  0.72603156]])

In [9]: W_non_nan = ~np.isnan(W)

In [10]: W_non_nan
Out[10]: 
array([[ True,  True],
       [ True,  True],
       [ True, False],
       [ True,  True],
       [False,  True],
       [ True,  True],
       [ True,  True],
       [ True,  True],
       [ True,  True],
       [ True,  True]], dtype=bool)

In [11]: A[W_non_nan]

Out[11]: 
array([ 0.60879191,  0.13556633,  0.23775325,  0.5744113 ,  0.44320766,
        0.81903997,  0.15170996,  0.11521106,  0.57292616,  0.47236245,
        0.1478904 ,  0.1597816 ,  0.6395053 ,  0.40529633,  0.25381449,
        0.28045816,  0.91827695,  0.85382925])

Barring A[2, 1] and A[4, 0]; the first two columns of A should be replaced with W. What's the cleanest way to do so?

Divakar · Accepted Answer · 2017-04-29 09:22:15Z

1

You can use masking/boolean-indexing with a mask of non-NaNs sliced into the first n' columns and use that mask for both selecting elements off W and assigning into the sliced portion of A, like so -

mask = ~np.isnan(W)
A[:,:mask.shape[1]][mask] = W[mask]

Sample run -

In [121]: A
Out[121]: 
array([[ 68.,  54.,  54.,  15.],
       [ 50.,  86.,  84.,  74.],
       [ 27.,  39.,  43.,  24.],
       [ 41.,  45.,  42.,  40.],
       [ 32.,  90.,  93.,  19.]])

In [122]: W
Out[122]: 
array([[  3.,   4.],
       [  3.,   1.],
       [  4.,  nan],
       [  6.,   8.],
       [ nan,   6.]])

In [123]: mask = ~np.isnan(W)

In [124]: A[:,:mask.shape[1]][mask] = W[mask]

In [125]: A
Out[125]: 
array([[  3.,   4.,  54.,  15.],
       [  3.,   1.,  84.,  74.],
       [  4.,  39.,  43.,  24.],
       [  6.,   8.,  42.,  40.],
       [ 32.,   6.,  93.,  19.]])

edited Apr 29, 2017 at 9:22

answered Apr 29, 2017 at 9:15

Divakar

222k19 gold badges273 silver badges374 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Nipun Batra Over a year ago

Thanks. Super useful!

Allen Qin · Accepted Answer · 2017-04-29 10:00:12Z

1

An alternative solution:

A = np.array([[ 68.,  54.,  54.,  15.],
       [ 50.,  86.,  84.,  74.],
       [ 27.,  39.,  43.,  24.],
       [ 41.,  45.,  42.,  40.],
       [ 32.,  90.,  93.,  19.]])

W = np.array([[  3.,   4.],
       [  3.,   1.],
       [  4.,  np.nan],
       [  6.,   8.],
       [ np.nan,   6.]])   
#replace first two cols of A with W except the positions where W has nan.    
A[:,0:2] = np.where(np.isnan(W),A[:,0:2],W)
print(A)

[[  3.   4.  54.  15.]
 [  3.   1.  84.  74.]
 [  4.  39.  43.  24.]
 [  6.   8.  42.  40.]
 [ 32.   6.  93.  19.]]

answered Apr 29, 2017 at 10:00

Allen Qin

20k9 gold badges55 silver badges68 bronze badges

Collectives™ on Stack Overflow

NumPy: Setting subset of numpy array at non-NaN positions

2 Answers 2

1 Comment

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

1 Comment

Comments

Your Answer

Sign up or log in

Post as a guest

Related