Categorical cross-entropy loss in TensorFlow - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to import the categorical cross-entropy loss function from TensorFlow.

TensorFlow

from tensorflow.keras.losses import [1]

Drag options to blanks, or click blank then click option'

Acategorical_crossentropy

Bmean_squared_error

Cbinary_crossentropy

Dhinge_loss

Attempts:

3 left

💡 Hint

Common Mistakes

Choosing binary_crossentropy which is for two-class problems.

Using mean_squared_error which is for regression tasks.

The categorical_crossentropy function is used for multi-class classification problems where labels are one-hot encoded.

2fill in blank

medium

Complete the code to compile a Keras model using categorical cross-entropy loss.

TensorFlow

model.compile(optimizer='adam', loss=[1], metrics=['accuracy'])

Drag options to blanks, or click blank then click option'

A'binary_crossentropy'

B'hinge'

C'mean_squared_error'

D'categorical_crossentropy'

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'binary_crossentropy' for multi-class problems.

Using 'mean_squared_error' which is for regression.

When compiling a model for multi-class classification with one-hot labels, use 'categorical_crossentropy' as the loss.

3fill in blank

hard

Fix the error in the code to compute categorical cross-entropy loss manually using TensorFlow.

TensorFlow

loss = tf.keras.losses.categorical_crossentropy(y_true, [1])

Drag options to blanks, or click blank then click option'

Ay_pred

Bpredictions

Cy_predicted

Dy_true

Attempts:

3 left

💡 Hint

Common Mistakes

Passing true labels twice instead of predictions.

Using a variable name that is not defined.

The function requires the true labels and the predicted probabilities as inputs. The predicted probabilities are usually stored in y_pred.

4fill in blank

hard

Fill both blanks to create a dictionary comprehension that maps class names to their predicted probabilities.

TensorFlow

class_probs = {cls: [1][i] for i, cls in enumerate([2])}

Drag options to blanks, or click blank then click option'

Ay_pred

By_true

Cclass_names

Dpredictions

Attempts:

3 left

💡 Hint

Common Mistakes

Using true labels instead of predicted probabilities.

Mixing up class names with predictions.

The dictionary comprehension pairs each class name with its predicted probability from y_pred. The class names are stored in class_names.

5fill in blank

hard

Fill all three blanks to compute the average categorical cross-entropy loss over a batch.

TensorFlow

losses = tf.keras.losses.categorical_crossentropy([1], [2])
avg_loss = tf.reduce_mean([3])

Drag options to blanks, or click blank then click option'

Ay_true

By_pred

Closses

Dpredictions

Attempts:

3 left

💡 Hint

Common Mistakes

Swapping true labels and predictions.

Trying to reduce mean over wrong variable.

First, compute the loss for each sample using true labels y_true and predictions y_pred. Then, calculate the average loss with tf.reduce_mean over the losses.