MLS-C01 Question 184
Select 3You are building a machine learning model using Amazon SageMaker and want to improve its performance by tuning hyperparameters. You decide to use SageMaker's built-in hyperparameter tuning capabilities. Which of the following steps are required to successfully configure and execute hyperparameter optimization in SageMaker?
- A
Define a range of hyperparameter values and specify them in a tuning job configuration.
- B
Select an appropriate objective metric to optimize during the tuning process.
- C
Manually iterate over different combinations of hyperparameters and update the model training script.
- D
Specify the maximum number of training jobs and parallel training jobs in the tuning job configuration.
- E
Use SageMaker Debugger to monitor training jobs and automatically stop poorly performing runs.
Show answer and explanation
Correct answers: A, B, D
Explanation
To perform hyperparameter optimization in SageMaker, you must define a range of hyperparameter values, select an objective metric to optimize, and configure the tuning job with settings like the maximum number of training jobs and parallel jobs. SageMaker handles the tuning process automatically, so manual iteration is not needed. Although SageMaker Debugger can add value to the process, it is not a core requirement for hyperparameter tuning.
- A. Correct.
Correct. Defining a range of hyperparameter values is required for SageMaker Hyperparameter Tuning to explore various combinations during optimization.
- B. Correct.
Correct. An objective metric must be defined so that SageMaker knows what to optimize (e.g., minimizing validation error or maximizing accuracy).
- C. Incorrect.
Incorrect. SageMaker's hyperparameter tuning jobs automate the process of exploring combinations, so manual iteration is not required.
- D. Correct.
Correct. Configuring the maximum number of training jobs and parallel training jobs is essential for controlling resource usage and tuning efficiency.
- E. Incorrect.
Incorrect. While SageMaker Debugger is useful for monitoring training jobs, it is not a mandatory step for hyperparameter optimization.