PROFESSIONAL-MACHINE-LEARNING-ENGINEER Question #11: Real Exam Question with Answer & Explanation

Question

You have a functioning end-to-end ML pipeline that involves tuning the hyperparameters of your ML model using AI Platform, and then using the best-tuned parameters for training. Hypertuning is taking longer than expected and is delaying the downstream processes. You want to speed up the tuning job without significantly compromising its effectiveness. Which actions should you take? (Choose two.)

Options

• ADecrease the number of parallel trials.
• BDecrease the range of floating-point values.
• CSet the early stopping parameter to TRUE.
• DChange the search algorithm from Bayesian search to random search.
• EDecrease the maximum number of trials during subsequent training phases.

Explanation

The objective is to reduce hypertuning time without significantly compromising model quality. Option C (set early stopping to TRUE) allows AI Platform to automatically halt individual trials that are not showing improvement, saving time on unpromising paths while keeping the most effective trials running to completion. Option E (decrease maximum trials in subsequent training phases) reduces the total number of evaluations in later phases where the best hyperparameter region is already narrowed down, cutting time without sacrificing the quality of the search. Option A (decrease parallel trials) would actually slow down the tuning by reducing concurrency. Option B (decrease floating-point range) could significantly compromise model quality by excluding good values. Option D (switch from Bayesian to random search) is less efficient-Bayesian search uses prior results to intelligently select next trials, so switching to random search would require more trials to find equivalent results, making the problem worse.

No community discussion yet for this question.