Hyperparameter Tuning MACHINE-LEARNING-ASSOCIATE Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   What is the defining characteristic of random search for hyperparameter tuning?

   • AIt starts with a grid and randomly perturbs the points
   • BIt uses a surrogate model to predict promising regions
   • CIt samples hyperparameters randomly from specified distributions or ranges
   • DIt evaluates all possible combinations of hyperparameter values
   Reveal Solution Discussion Select an answer first
2. 2foundation · easy

   What does the `parallelism` parameter in `SparkTrials` control?

   • AThe total number of trials to run in the optimization
   • BThe maximum number of trials that can run concurrently
   • CThe number of Spark executors used for each trial
   • DThe number of hyperparameters to tune in parallel
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   What must a Hyperopt objective function return?

   • AThe trained model object
   • BA dictionary containing the loss and the hyperparameters used
   • CThe accuracy of the model on the training set
   • DA single numeric loss value to minimize
   Reveal Solution Discussion Select an answer first
5. 4foundation · easy

   What is a common way to monitor the progress of a parallel Hyperopt tuning run?

   • AWait until the run completes and then review the final results
   • BCheck the Spark UI for the number of completed tasks
   • CUse the `trials` object to inspect the results of completed trials
   • DLook at the driver logs for each trial's output
   Reveal Solution Discussion Select an answer first
6. 5foundation · easy

   In a Hyperopt objective function, which of the following is the typical role of the input parameter (e.g., `params`)?

   • AIt is the loss value from the previous trial
   • BIt is a dictionary of hyperparameter values sampled from the search space
   • CIt is the maximum number of evaluations
   • DIt contains the training data and labels
   Reveal Solution Discussion Select an answer first