A team is tuning hyperparameters for a classification model. They have a small dataset and are concerned about overfitting to the validation set during tuning. They want to get a reliable estimate of model performance for each hyperparameter configuration. Which approach should they use?

Use k-fold cross-validation to evaluate each configuration and select the one with the best average validation score

Use a single hold-out validation set and tune hyperparameters to maximize performance on it

Use a separate test set for tuning and then evaluate on the validation set

Use the training set accuracy to select hyperparameters, since it reflects the model's true performance

Why is random search often more efficient than grid search in high-dimensional hyperparameter spaces?

It requires fewer training runs because it stops early when performance plateaus

It guarantees finding the optimal hyperparameters

It uses a probabilistic model to focus on promising areas

It does not evaluate all combinations, so it can explore more values per hyperparameter with the same budget

Which benefit is most associated with Bayesian optimization compared to grid search?

It is simpler to implement than grid search

It typically requires fewer training runs to find a good hyperparameter configuration

It guarantees finding the global optimum

It works only with continuous hyperparameters

A team has a k-nearest neighbors (k-NN) model that is too slow at inference time because it computes distances to all training samples for each prediction. They want to reduce runtime while maintaining accuracy. Which structural modification is most appropriate?

Increase the value of k to reduce the number of neighbors considered

Use a KD-tree or ball tree data structure to accelerate nearest neighbor search

Use a larger training set to improve accuracy

Switch to a brute-force approach to simplify the algorithm

When optimizing an algorithm, what is a common trade-off to consider?

Training time versus the size of the test set

Computational resources versus the programming language used

Model performance versus training time and computational resources

Model performance versus the number of features in the dataset

Certified Artificial Intelligence Practitioner (CAIP)

Domain 3Objective 2

Objective 3.2 Optimize the Algorithm (e.g., Structure, Run Time, Tuning Hyperparameters) AIP-210 Practice Questions (Page 4)

Part of the 3.0 Training and Tuning ML Systems and Models domain, which accounts for 24% of the AIP-210 exam.

30questions here

6free pages

9concepts

24%of the exam

Questions 16–20

16application · medium
A team is tuning hyperparameters for a classification model. They have a small dataset and are concerned about overfitting to the validation set during tuning. They want to get a reliable estimate of model performance for each hyperparameter configuration. Which approach should they use?
Select an answer first
17foundation · easy
Why is random search often more efficient than grid search in high-dimensional hyperparameter spaces?
Select an answer first
18foundation · easy
Which benefit is most associated with Bayesian optimization compared to grid search?
Select an answer first
19application · medium
A team has a k-nearest neighbors (k-NN) model that is too slow at inference time because it computes distances to all training samples for each prediction. They want to reduce runtime while maintaining accuracy. Which structural modification is most appropriate?
Select an answer first
20foundation · easy
When optimizing an algorithm, what is a common trade-off to consider?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to AIP-210

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by CertNexus. “AIP-210” is a trademark of its owner, used for identification only.