CODEWITHSUNDEEP
python-3.xtensorflowmachine-learning
Taimur Islam
Taimur Islam

Reputation: 990

Categorical level to one hot encoding in python tensorflow

If i have a categorical label like this

labels = [cat,dog, bird, cow]

now i want to convert it like one hot encoding. Is it possible by using tensorflow. like this

output_label = [[1 0 0 0]
               [0 1 0 0]
               [0 0 1 0]
               [0 0 0 1]]

Upvotes: 1

Views: 979

Answers (1)

Tim
Tim

Reputation: 10709

First you need to transform your categorical data to numerical format. You could do that for example like this:

def categorical_to_numerical(labels):
    num_labels=[]
    for k in labels:
        if k == 'cat':
            num_labels.append(0)
        if k == 'dog':
            num_labels.append(1)
        if k == 'bird':
            num_labels.append(2)
        if k == 'cow':
            num_labels.append(3)
    return num_labels

print labels
// prints ['cat','dog', 'bird', 'cow', 'dog', 'bird'] 
print categorical_to_numerical(labels)
// prints [0, 1, 2, 3, 1, 2]

Now you can easily use a tensorflow built-in function called tf.one_hot:

indices = categorical_to_numerical(labels)
detph = 4 // because you have four categories 
one_hot_labels = tf.one_hot(indices, depth)

Read more about tf.one_hot here.

Upvotes: 1

Related Questions