I have the following array:
myArray1 = np.array([[255, 255, 255, 1, 1, 255, 255, 255],
[255, 1, 1, 1, 1, 1, 255, 255]])
I want to create a new array myArray2 based on myArray1. Where myArray1 has a value of 1, I want to replace the value by numpy.random.normal(loc=40, scale=10).
How can I do that?
If your array is called a, you can do
b = a == 1
a[b] = numpy.random.normal(loc=40, scale=10, size=b.sum())
This avoids Python loops and takes advantage of Numpy's performance.
The first line creates a Boolean array with the same shape as a that is True in all entries that correspond to ones in a. The second line uses advanced indexing with this Boolean array to assign new values only to the places where b is True.
ValueError: array is not broadcastable to correct shape
How about a list comprehension like this:
[i if i != 1 else numpy.random.normal(loc=40, scale=10) for j in myArray1 for i in j]
j loops through each of the rows of myArray1 and i loops through each element of the row.
