Designing and Implementing a Data Science Solution on Azure — Question 178

You are implementing hyperparameter tuning by using Bayesian sampling for an Azure ML Python SDK v2-based model training from a notebook. The notebook is in an Azure Machine Learning workspace. The notebook uses a training script that runs on a compute cluster with 20 nodes.

The code implements Bandit termination policy with slack_factor set to 0.2 and a sweep job with max_concurrent_trials set to 10.

You must increase effectiveness of the tuning process by improving sampling convergence.

You need to select which sampling convergence to use.

What should you select?

Answer options

• A. Set the value of max_concurrent_trials to 20.
• B. Set the value of slack_factor of early_termination policy to 0.1.
• C. Set the value of slack_factor of early_termination policy to 0.9.
• D. Set the value of max_concurrent_trials to 4.

Correct answer: D

Explanation

Option D is correct because reducing the number of concurrent trials can lead to better convergence in sampling by allowing more focused exploration of hyperparameter space. Increasing max_concurrent_trials (A) would lead to less effective sampling as it spreads the trials too thinly. Changing the slack_factor (B and C) affects the termination policy but does not directly enhance the convergence efficiency in this scenario.