DP-100 Question #523: Real Exam Question with Answer & Explanation

Question

You have an Azure Machine Learning workspace. You plan to implement automated hyperparameter tuning for model training in the workspace. You need to select the sweep jobs parameter sampling method that will randomize the selection of hyperparameters from the search space but allow for reproducing search results. Which sampling method should you use?

Options

• ASobol
• BGrid
• CRandom
• DBayesian

Explanation

{"question_number": 2, "correct_answer": "A", "explanation": "Sobol sampling is a quasi-random, low-discrepancy sequence method. Unlike true random sampling, Sobol uses a deterministic mathematical sequence that covers the search space evenly while remaining reproducible - running the sweep again with the same configuration produces the same sequence of hyperparameter combinations. Grid search is exhaustive (not randomized). Random sampling is non-deterministic and cannot reproduce exact results without a fixed seed. Bayesian sampling uses prior results to guide the next selection, making it adaptive rather than reproducible in the same sense.", "generated_by": "claude-sonnet", "llm_judge_score": 4}

No community discussion yet for this question.