Answer-first summary for fast verification
Answer: Set slack_factor to 0.2.
The question requires configuring a Bandit termination policy where a primary metric of 0.8 should terminate runs below 0.66. According to the Microsoft documentation referenced in the community discussion, the Bandit policy uses the formula: threshold = best_metric / (1 + slack_factor). Given best_metric = 0.8 and threshold = 0.66, solving 0.66 = 0.8 / (1 + slack_factor) gives slack_factor = (0.8 / 0.66) - 1 ≈ 0.212, which rounds to 0.2. The community consensus (100% votes for C, upvoted comments with documentation links) confirms that setting slack_factor to 0.2 achieves this outcome. Options A and D (slack_amount) are incorrect as slack_amount uses an absolute difference (best_metric - slack_amount), not a ratio. Option B (slack_factor = 0.1) would set the threshold to 0.8 / 1.1 ≈ 0.727, which is higher than 0.66 and does not match the requirement.
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
You create an Azure Machine Learning workspace and are implementing hyperparameter tuning for a model training run from a notebook.
You must configure a Bandit termination policy with the following behavior: If the primary metric (AUC) is 0.8 at the evaluation intervals, any run where the primary metric falls below 0.66 should be terminated.
Which Bandit termination policy configuration should you use?
A
Set slack_amount to 0.2.
B
Set slack_factor to 0.1.
C
Set slack_factor to 0.2.
D
Set slack_amount to 0.1.
No comments yet.