Keras - One-Hot Encoding

Introduction

In this tutorial, we want to one-hot encode a NumPy array that contains categorical values. In order to do this, we use the to_categorical() function of Keras.

Import Libraries

First, we import the following python modules:

import numpy as np
from keras.utils import to_categorical

Define Data

Next, we create a NumPy array that contains the categorical values 0, 1 and 2.

x = np.array([1, 2, 0, 0, 2, 1, 2, 1, 1, 0])

print(x)

One-Hot Encoding

Now, we want to transform categorical numerical values into binary vectors. The principle of One-Hot Endoding is visualized in the following graphic: