[FIXED] How do I scale each element of a numpy array relative to itself?

Issue

I'm trying to apply MinMaxScaler to each element of a numpy array, and I want it vectorized (I don't want to use a for loop).

example = np.array([[2.52163839, 2.54165282, 2.12608389, 2.54515915],
                    [2.29481214, 1.78448378, 2.26652405, 2.27311454],
                    [2.31706137, 2.29058921, 1.83225955, 2.29767736]])

I want the first element ([2.52163839, 2.54165282, 2.12608389, 2.54515915]) scaled relative to itself (so index 2 becomes 0, index 3 becomes 1, etc.),

then the second element ([2.29481214, 1.78448378, 2.26652405, 2.27311454]) scaled relative to itself (so index 1 becomes 0, index 0 becomes 1, etc.), and so on for the whole array.

I tried example = np.array(list(map(MinMaxScaler().fit_transform(), example))), but it doesn't work because MinMaxScaler requires the thing being scaled to be passed as an argument into the fit_transform method.

Thanks!

Solution

One can use map with a lambda function. Some reshaping will be needed due to the specifications of MinmaxScaler

np.array(list( map ( lambda x: MinMaxScaler().fit_transform(\
  x.reshape(-1, 1)) , example) )).reshape(3,4)

Output :

array([[0.94387462, 0.99163317, 0.        , 1.        ],
       [1.        , 0.        , 0.94456885, 0.95748306],
       [1.        , 0.94539591, 0.        , 0.96001663]]) 

==Edit: Addition to include @NaiveBae's own solution==

Simply scale a Transposed array:

MinMaxScaler().fit_transform( example.T ).T

This will work because MinMaxScaler uses the following as its implementation:

X_std = (X - X.min(axis=0)) / (X.max(axis=0) - X.min(axis=0))
X_scaled = X_std * (max - min) + min

By transposing , we convert our required axis to axis 0 (first axis).

Answered By - R.S.