Is there a way to OneHotEncode an entire array?

[moisestohias]

Hi,
I wonder, Is there a way to OneHotEncode an entire array?

def OneHot(y, NumOfClasses):
    e = np.zeros((NumOfClasses, 1))
    e[y] = 1.0
    return e

Basically obviating doing this np.array([OneHot(i, 10) for i in y]),
I tried using np.vectorize function, but that didn't work because of indexing issues I guess, I am not sure.
Anyway I would appreciate you suggests.

[mrdbourke]

Hi @moisestohias,

You may want to try Scikit-Learn's OneHotEncoder - http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.OneHotEncoder.html

[moisestohias]

I have found many solutions on the Internet
Different way to one hot

Resources:

• stackoverflow:1
• stackoverflow:2

!Note: for one hot encoding you can use scipy.sparse as well.
!TIP: don't rely on the a.max()+1 to get the number of classes.

__Note: using eye is the only useful solution for an input N-D matrix to one-hot N+1D matrix. __
Example:

input_matrix=np.asarray([[0,1,1] , [1,1,2]]) ; np.eye(3)[input_matrix] # output 3D tensor
def one_hot(a, classes=10): return np.eye(classes)[a] # .T Transpose if you don't want torch style

def one_hot(a: np.ndarray):
  # a.max()+1 automatically determin the number of classes based on the highest value.
  hot = np.zeros((a.size, a.max()+1), dtype=np.float32)
  hot[np.arange(a.size),a] = 1.0
  return hot

def one_hot(a, classes=10 ):
  hot = np.zeros((a.size, classes), dtype=np.float32)
  hot[np.arange(a.size),a] = 1.0
  return hot