MLS-C01 Question 190
Select 3You are training a deep learning model on AWS SageMaker to classify images, but the model is overfitting the training data. To address this issue, you decide to implement dropout in the model architecture. Which of the following statements about dropout are correct?
- A
Dropout randomly deactivates a fraction of neurons during training to improve generalization.
- B
Dropout should only be applied during the inference phase to avoid underfitting.
- C
Dropout helps prevent the model from relying too heavily on specific neurons during training.
- D
Increasing the dropout rate to 100% will typically result in better model performance.
- E
Dropout is typically disabled during the evaluation or inference phase.
Show answer and explanation
Correct answers: A, C, E
Explanation
Dropout is a regularization technique used to reduce overfitting in neural networks. During training, it randomly deactivates a fraction of neurons in each layer, encouraging the model to learn more generalized features. However, dropout is disabled during inference or evaluation to ensure the full capacity of the model is used for predictions. Misusing dropout, such as applying it during inference or setting the dropout rate too high, can lead to poor model performance.
- A. Correct.
Correct: Dropout works by randomly deactivating a fraction of neurons during training, which helps improve generalization by preventing co-adaptation of neurons.
- B. Incorrect.
Incorrect: Dropout is only applied during training to prevent overfitting. During inference, all neurons are used to make predictions.
- C. Correct.
Correct: Dropout forces the model to learn more robust features by preventing reliance on specific neurons.
- D. Incorrect.
Incorrect: Setting the dropout rate to 100% deactivates all neurons, which prevents the model from learning anything.
- E. Correct.
Correct: Dropout is designed to be disabled during inference or evaluation, as all neurons should be used to make predictions.