I am creating a numpy array using the following:
X = np.linspace(-5, 5, num=500)
This generates points evenly sampled 500 points between -5 and 5. The shape of the resulting array is: (500,). Now, I need to pass it to a function that expects a 2-D array. So, I can reshape it as:
X = X.reshape((500, 1))
However, I noticed that X = X[:, None] has the same effect. For the life of me though, I cannot understand what this syntax is doing. Hoping someone can shed some light on this.
The syntax X[: ,None] is actually the same as:
X[:, np.newaxis]
which is adding a new dimension to your original array.
The Python interpreter translates
x[:,None]
to
x.__getitem__((slice(None,None,None), None))
and the ndarray implementation of __getitem__ acts in much the same way as x.reshape(500,1). Implementation details will differ, but the effect is the same. `
So at a syntax level, it's just normal Python. But the numpy semantics give it a distinctive meaning.
x[:, np.newaxis]
may be clearer, but np.newaxis is just an alias for None:
In [48]: np.newaxis is None
Out[48]: True
Noneisnp.newaxis- this page on advanced indexing explains its use in a bit of detail and might be a useful reference.