MLA-C01 Question #63: Real Exam Question with Answer & Explanation

Question

A company is building a deep learning model on Amazon SageMaker. The company uses a large amount of data as the training dataset. The company needs to optimize the model's hyperparameters to minimize the loss function on the validation dataset. Which hyperparameter tuning strategy will accomplish this goal with the LEAST computation time?

Options

• AHyperband
• BGrid search
• CBayesian optimization
• DRandom search

Explanation

Hyperband is correct because it uses an early-stopping mechanism called "successive halving" - it runs many configurations simultaneously, quickly terminates the worst-performing ones, and allocates more compute budget to promising candidates. This aggressive pruning makes it significantly faster than alternatives when working with large datasets and many hyperparameter combinations.

Grid search (B) exhaustively tries every combination of hyperparameter values, which is computationally prohibitive at scale - it's the slowest option. Bayesian optimization (C) is smarter than grid/random search by building a probabilistic model to guide the next trial, but it runs trials sequentially and doesn't eliminate poor performers early, making it slower than Hyperband for large-scale problems. Random search (D) samples hyperparameter configurations randomly and is better than grid search, but wastes compute on bad configurations because it has no early-stopping mechanism.

Memory tip: Think of Hyperband as a tournament bracket - it runs many competitors at once, quickly knocks out the losers, and only invests resources in the finalists. "Hyperbandwidth" = fastest because it cuts waste early.

No community discussion yet for this question.