Question

A data scientist is tuning an XGBoost model with SageMaker Automatic Model Tuning. The search space includes 6 continuous hyperparameters, and the team has budgeted a maximum of 40 training jobs. They want to find the best-performing combination as efficiently as possible within this limited job budget. Which tuning strategy should they configure for the hyperparameter tuning job?

Accepted Answer

Bayesian optimization, which uses results from prior jobs to inform which hyperparameter combinations to try next

Answer

Grid search, which exhaustively evaluates every combination of hyperparameter values

Answer

Random search, which samples hyperparameter combinations independently and in parallel

Answer

Hyperband, which relies on evaluating many low-fidelity partial runs of a fixed grid