AWS ML Model Training and Tuning Quiz

1. What does the learning rate control in model training?

Step size for weight updates during backpropagation

Number of training epochs

Batch size per iteration

Model prediction speed

2. Which metric is most appropriate for evaluating imbalanced classification models?

Accuracy

F1 score or AUC-ROC

Training time

Model size

3. What is the benefit of using spot instances for SageMaker training jobs?

Guaranteed training completion

Significant cost reduction with interruption handling

Faster model convergence

Automatic hyperparameter selection

4. In SageMaker training, what does the checkpoint mechanism allow?

Resume interrupted training from saved state

Skip validation steps

Reduce model accuracy

Eliminate hyperparameter tuning

5. Which technique reduces variance in model predictions by combining multiple models?

Dropout

Ensemble methods

Batch normalization

Weight decay

6. What does SageMaker Debugger monitor during training?

Only model accuracy

Training metrics, system metrics, and model behavior

Instance costs only

Data preprocessing logs

7. In SageMaker, which built-in algorithm supports automatic model tuning?

Linear Learner, XGBoost, and Neural networks

Only custom algorithms

Manual tuning only

No algorithms support automatic tuning

8. What is data augmentation used for in model training?

Increase dataset size and improve generalization

Reduce training time

Eliminate preprocessing steps

Improve prediction speed

9. Which metric should you monitor to detect overfitting during training?

Training loss only

Validation loss diverging from training loss

Number of epochs

Instance memory usage

10. In SageMaker, what is the role of the warm pool in Hyperparameter Tuning?

Store completed training jobs for reuse

Reduce tuning costs by reusing instances

Increase data processing speed

Automatically scale training instances

11. Which AWS service allows you to automatically tune hyperparameters for SageMaker training jobs?

SageMaker Hyperparameter Tuning

AWS Lambda

Amazon CloudWatch

AWS Glue

12. Which SageMaker feature enables efficient training with very large datasets?

SageMaker Experiments

SageMaker Pipelines

Pipe mode for streaming data

SageMaker Debugger

13. What is the purpose of a validation set during model training?

To train the model

To evaluate performance and tune hyperparameters

To preprocess data

To deploy the model

14. Which sampling strategy is used in SageMaker Hyperparameter Tuning to explore the hyperparameter space?

Grid search only

Random search, grid search, and Bayesian optimization

Manual tuning only

No search strategy needed

15. In SageMaker, what does batch normalization help achieve?

Faster convergence and improved stability

Reduce model size

Eliminate hyperparameter tuning

Increase training data requirements

16. What is early stopping in the context of neural network training?

Stopping training before convergence

Halting when validation loss stops improving

Terminating the training instance

Reducing the learning rate

17. Which technique helps prevent overfitting during model training?

Increasing batch size

Regularization (L1/L2)

Using only training data

Removing validation sets

18. What does distributed training in SageMaker allow you to do?

Train only on CPU instances

Parallelize training across multiple instances