You create an Azure Machine Learning workspace. You are implementing hyperparameter tuning for a model training from a notebook. You must configure a Bandit termination policy that provides the…

Question

Options

• ASet slack_amount to 0.2.
• BSet slack_factor to 0.1.
• CSet slack_factor to 0.2.
• DSet slack_amount to 0.1.

• A
  15% (4)
• B
  4% (1)
• C
  70% (19)
• D
  11% (3)

Explanation

The Bandit early termination policy terminates runs whose primary metric falls below a threshold derived from the best run so far. With slack_factor, the termination threshold is calculated as: best_metric / (1 + slack_factor). Plugging in slack_factor = 0.2: 0.8 / (1 + 0.2) = 0.8 / 1.2 ≈ 0.667, which matches the required cutoff of 0.66. With slack_amount, the threshold is best_metric − slack_amount, so slack_amount = 0.2 gives 0.8 − 0.2 = 0.6 (too low), and slack_amount = 0.1 gives 0.8 − 0.1 = 0.7 (too high). Only slack_factor = 0.2 produces the correct ~0.66 threshold.

No community discussion yet for this question.