Question 2
Domain 3: Model DevelopmentA data scientist is tuning hyperparameters for a model that runs on a single machine and wants to reduce total tuning time without changing the model itself. Which approach is most appropriate?
Correct answer: B
Explanation
When a model runs on a single node, hyperparameter tuning can be accelerated by evaluating multiple hyperparameter configurations in parallel on that same machine. — official.txt
Why each option is right or wrong
A. Move the model to a distributed multi-node architecture before tuning any parameters
Parallel processing is used to speed tuning for models already running on a single node.
B. Run multiple hyperparameter tuning trials in parallel on the single node
The source states that parallel processing is used to speed up hyperparameter tuning for models running on a single node. In this scenario, the goal is to reduce tuning time without changing the model, so evaluating multiple hyperparameter configurations in parallel on that machine directly matches the stated use case.
C. Reduce the number of hyperparameters so tuning no longer needs parallel execution
Reducing hyperparameters may change the search, but it is not the stated method for speeding tuning.
D. Train each model sequentially and compare results after all runs are complete
Sequential training does not use parallel processing to speed hyperparameter tuning.