フリー問題

問題文

You must tune three hyperparameters and have budget for 30 training jobs. Which sampling algorithm is recommended for this budget, and why?

選択肢

1. Grid sampling, because 30 jobs are enough to cover any search space exhaustively
2. Bayesian sampling, because the recommended budget is at least 20 times the number of hyperparameters being tuned
3. No sampling is needed because automated machine learning tunes hyperparameters without a sweep
4. Random sampling with the maximum number of concurrent jobs, because more parallelism always improves the final result