Why Numpy omits the dimension of an array-CodePudding

I have the following problem, I have an array that contains 60 arrays, inside them there are 21 arrays, and inside these last ones there are 3 integers, when I do the shape, it says (60,) How can I make the reshape to be (60,21,3)?

CodePudding user response：

I am not entirely sure what you are asking, you didn't leave a whole lot of details. However, here are some basic numpy functions that seem like they would answer your question.

arr = np.random.rand(2,2,1)
print(arr)
#obviously you can use your dimensions of (60,21,3)
#however I am going to print the array

output:

array([[[0.76494962],
        [0.20654063]],

       [[0.22742839],
        [0.15418584]]])

If you want to print the exact shape, in this case of (2,2,1) print(arr.shape) gives the output (2, 2, 1). This is a tuple object stored within the numpy array, therefore will not change unless you re-define the array with a new numpy array. If numpy does not output (60,21,3) in your case, there is a very good chance you set your code up incorrectly.

If I want to resize my array it is important that you adhere to the size of your array, in my case if I use the function np.size(arr) I get 4. This is because 2*2*1 = 4 so I must reshape it to a size of something that adheres to the same size. for example these are valid calls:

arr = arr.reshape(4)
arr = arr.reshape(1,1,4)
arr = arr.reshape(1,1,2,2)

And so on. As long as if you multiply the numbers they come out to np.size(arr). In your case you have the shape (60,21,3) therefore you must reshape it to have the size 60*21*3 = 3780.

I also left an example of the resize function below. Again, you did not leave much in terms of details on what you wanted, so I am assuming it is one of the functions or explanations I have outlined in this post.

>>> import numpy as np
>>> arr = np.random.rand(60,21,3)
>>> arr.shape
(60, 21, 3)
>>> arr = arr.reshape(3780)
>>> arr.shape
(3780,)
>>> arr = np.random.rand(60)
>>> arr.shape
(60,)
>>> arr = np.resize(arr, (60,21,3))
>>> arr.shape
(60, 21, 3)

CodePudding user response：

You probably created a array of lists (objects), for example:

b = np.array([0], dtype=np.dtype(object))
b[0] = [4,5,6]
b.shape

Gives b as: array([list([4, 5, 6])], dtype=object)

and the shape will be (1,), not 'seeing' the 2nd dimension represented by the list

Check of this is the case by running b.dtype