In neural network training, what is the primary purpose of dropout?

To add noise to the input data to make the model more robust

To reduce the learning rate during training to improve convergence

To randomly deactivate a fraction of neurons during each training iteration to prevent overfitting

To remove outliers from the training dataset

A deep learning engineer is training a recurrent neural network for time series forecasting. The training loss is decreasing very slowly, and the gradients are very small. The engineer suspects the vanishing gradient problem. Which optimization strategy is most appropriate?

Increase the learning rate to amplify the small gradients.

Switch from SGD to Adam, which uses adaptive learning rates for each parameter.

Add L2 regularization to the weights to prevent them from becoming too small.

Use a different activation function, such as ReLU, which has a constant gradient of 1 for positive inputs.

A machine learning team is tuning hyperparameters for a gradient boosting model for anomaly detection. The model has a large number of hyperparameters, including learning rate, number of estimators, max depth, and subsample. The team has a limited compute budget and wants to find the best configuration. They also need to avoid overfitting. Which approach is most appropriate?

Use grid search with cross-validation, exploring all combinations of a small set of values.

Use random search with cross-validation, sampling a fixed number of random combinations.

Use manual tuning, starting with the team's best guess and adjusting based on validation performance.

Use Bayesian optimization with cross-validation, using a surrogate model to guide the search.

A team is training a large transformer model for natural language processing. The model overfits the training data, and the team has limited GPU time. They need to reduce overfitting while keeping the training time manageable. Which combination of techniques is most appropriate?

Train for more epochs with a lower learning rate.

Increase the model size and use a higher learning rate.

Apply dropout and early stopping.

GIAC Machine Learning Engineer

Domain 3Objective 1

Anomaly Detection and Optimization GMLE Practice Questions (Page 5)

Part of the Advanced Machine Learning and Neural Networks domain, which makes up ~30% of our current practice bank. GIAC (SANS) does not publish an official question count, but from its 180-minute exam (~70–120 total, ~21–36 in this domain), expect 7–12 from this objective — we provide 47 practice questions to prepare you well beyond it. (estimate)

47questions here

10free pages

9concepts

Questions 21–25

21foundation · easy
In neural network training, what is the primary purpose of dropout?
Select an answer first
22application · medium
A deep learning engineer is training a recurrent neural network for time series forecasting. The training loss is decreasing very slowly, and the gradients are very small. The engineer suspects the vanishing gradient problem. Which optimization strategy is most appropriate?
Select an answer first
23expert · hard
A machine learning team is tuning hyperparameters for a gradient boosting model for anomaly detection. The model has a large number of hyperparameters, including learning rate, number of estimators, max depth, and subsample. The team has a limited compute budget and wants to find the best configuration. They also need to avoid overfitting. Which approach is most appropriate?
Select an answer first
24foundation · easy
Which optimization algorithm is specifically designed to address the problem of vanishing gradients in deep neural networks by using adaptive learning rates for each parameter?
Select an answer first
25expert · hard
A team is training a large transformer model for natural language processing. The model overfits the training data, and the team has limited GPU time. They need to reduce overfitting while keeping the training time manageable. Which combination of techniques is most appropriate?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to GMLE

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by GIAC (SANS). “GMLE” is a trademark of its owner, used for identification only.

Anomaly Detection and Optimization GMLE Practice Questions (Page 5)

Questions 21–25

In neural network training, what is the primary purpose of dropout?

A deep learning engineer is training a recurrent neural network for time series forecasting. The training loss is decreasing very slowly, and the gradients are very small. The engineer suspects the vanishing gradient problem. Which optimization strategy is most appropriate?

Which optimization algorithm is specifically designed to address the problem of vanishing gradients in deep neural networks by using adaptive learning rates for each parameter?

A team is training a large transformer model for natural language processing. The model overfits the training data, and the team has limited GPU time. They need to reduce overfitting while keeping the training time manageable. Which combination of techniques is most appropriate?

Keep studying — Advanced Machine Learning and Neural Networks