Tune Model Hyperparameters SNOWPRO-ADVANCED-DATA-SCIENTIST Practice Questions (Page 7)

Questions 31–32

1. 31expert · hard

   A data science team is tuning a model with a very expensive evaluation function (each evaluation takes 1 hour). They have a budget of 24 hours. They need to find a good configuration with as few evaluations as possible. Which approach is most appropriate?

   • AGrid search with 5 values per hyperparameter (3 hyperparameters)
   • BManual tuning with 10 runs
   • CRandom search with 24 iterations
   • DBayesian optimization with 24 trials
   Reveal Solution Discussion Select an answer first
2. 32foundation · easy

   Which of the following is a primary disadvantage of grid search compared to random search?

   • AIt requires more data than random search to achieve the same performance
   • BIt may miss optimal values that lie between the predefined grid points
   • CIt always overfits the validation set
   • DIt cannot be used with continuous hyperparameters
   Reveal Solution Discussion Select an answer first