Python – tensorflow.boolean_mask() method

TensorFlow is open-source Python library designed by Google to develop Machine Learning models and deep learning  neural networks. boolean_mask() is method used to apply boolean mask to a Tensor.

Syntax: tensorflow.boolean_mask(tensor, mask, axis, name)

Parameters:

• tensor: It’s a N-dimensional input tensor.
• mask: It’s a boolean tensor with k-dimensions where k<=N and k  is known statically.
• axis: It’s a 0-dimensional tensor which represents the axis from which mask should be applied. Default value for axis is zero and k+axis<=N.
• name: It’s an optional parameter that defines the name for the operation.

Return: It returns (N-K+1)-dimensional tensor which have the values that are populated against the True values in mask. 

Example 1: In this example input is 1-D.

Python3

# importing the library
import tensorflow as tf

# initializing the inputs
tensor = [1,2,3]
mask = [False, True, True]

# printing the input
print('Tensor: ',tensor)
print('Mask: ',mask)

# applying the mask
result = tf.boolean_mask(tensor, mask)

# printing the result
print('Result: ',result)

Output:

Tensor:  [1, 2, 3]
Mask:  [False, True, True]
Result:  tf.Tensor([2 3], shape=(2,), dtype=int32)

Example 2: In this example 2-D input is taken.

Python3

# importing the library
import tensorflow as tf

# initializing the inputs
tensor = [[1, 2], [10, 14], [9, 7]]
mask = [False, True, True]

# printing the input
print('Tensor: ',tensor)
print('Mask: ',mask)

# applying the mask
result = tf.boolean_mask(tensor, mask)

# printing the result
print('Result: ',result)

Output:

Tensor:  [[1, 2], [10, 14], [9, 7]]
Mask:  [False, True, True]
Result:  tf.Tensor(
[[10 14]
 [ 9  7]], shape=(2, 2), dtype=int32)