Defining a TensorFlow constant Tensor for Neural Network Weights

Python TensorFlow Building and Training a Simple Model: Exercise-2 with Solution

Write a Python program that defines a TensorFlow constant tensor containing weights for a neural network layer with shape (10, 4).

Sample Solution:

Python Code:

import tensorflow as tf

# Define the shape of the weight tensor
weight_shape = (10, 4)

# Create a TensorFlow constant tensor for weights
weights = tf.constant(tf.random.normal(weight_shape), dtype=tf.float32)

# Print the weight tensor
print("Weight Tensor:")
print(weights)

Output:

Weight Tensor:
tf.Tensor(
[[-0.17399755 -0.06166992 -0.18227974  0.9468848 ]
 [ 0.6667386   0.7858261   1.5395325  -0.448641  ]
 [ 0.8488986   0.06063127  0.50501484 -0.20491312]
 [-0.5653967  -0.2263971  -0.25578663  0.6188185 ]
 [-0.5225996   1.1885293   0.14677407 -1.5749815 ]
 [-0.59405106 -0.7491368   0.6874295  -0.23348485]
 [ 2.1346924   2.0820951   0.4871456  -1.4516023 ]
 [-1.5244085   0.9741064  -1.4729421  -1.0187328 ]
 [-1.1803584   0.00808251  0.36181325 -0.6317905 ]
 [-1.1563174  -1.4196043   1.0241832   1.4271173 ]], shape=(10, 4), dtype=float32)

Explanation:

In the exercise above -

• Import TensorFlow as tf.
• Define the weight tensor shape as (10, 4) to indicate that you want a tensor with 10 rows (input units) and 4 columns (output units).
• Use tf.random.normal(weight_shape) to create a random tensor with the specified shape.
• Specify the data type of the weights as tf.float32.
• Finally, we print the weight tensor to see its values.

Python Code Editor:

Previous: Creating a TensorFlow placeholder with variable batch size.
Next: Performing matrix multiplication with TensorFlow in Python.