How do overfitting and underfitting differ in terms of model performance on training data?

Both result in high training performance.

Both result in low training performance.

Overfitting results in high training performance, while underfitting results in low training performance.

Overfitting results in low training performance, while underfitting results in high training performance.

Which of the following is an effective way to reduce overfitting?

Using a smaller training dataset.

Using cross-validation to tune hyperparameters.

Increasing the number of model parameters.

Removing regularization from the model.

In a hands-on exercise, a student must deliberately create an overfitting model for a small dataset of 50 samples. Which setup is most likely to produce the desired result?

A decision tree with depth 2 trained on all 50 samples.

A degree-2 polynomial trained on 5 samples with early stopping.

A linear model trained on all 50 samples with strong L2 regularization.

A high-degree polynomial (degree 15) trained on all 50 samples with no regularization.

Which technique is commonly used to mitigate overfitting?

Increasing the model's complexity.

Training the model for more epochs.

Adding L2 regularization to the model's loss function.

Removing features from the dataset.

A team has trained a deep neural network that achieves 100% training accuracy but 70% test accuracy on a binary classification task. They have already tried increasing the amount of training data, but the test accuracy only improved to 72%. Which additional approach is most likely to yield the largest improvement in test accuracy?

Remove the dropout layers to allow the network to fit the data better.

Increase the learning rate to escape local minima.

Apply dropout and batch normalization, and reduce the network's capacity.

Use a smaller batch size to introduce more gradient noise.

Certified Tester AI Testing

Domain 3Objective 9

Hands-On Exercise: Demonstrate Overfitting and Underfitting CT-AI Practice Questions (Page 1)

Part of the Domain 3: Machine Learning (ML) - Overview domain, which makes up ~15% of our current practice bank. ISTQB does not publish an official question count, but from its 60-minute exam (~25–40 total, ~4–6 in this domain), expect 1–1 from this objective — we provide 26 practice questions to prepare you well beyond it. (estimate)

26questions here

6free pages

11concepts

Questions 1–5

1foundation · easy
How do overfitting and underfitting differ in terms of model performance on training data?
Select an answer first
2foundation · easy
Which of the following is an effective way to reduce overfitting?
Select an answer first
3application · medium
In a hands-on exercise, a student must deliberately create an overfitting model for a small dataset of 50 samples. Which setup is most likely to produce the desired result?
Select an answer first
4foundation · easy
Which technique is commonly used to mitigate overfitting?
Select an answer first
5expert · hard
A team has trained a deep neural network that achieves 100% training accuracy but 70% test accuracy on a binary classification task. They have already tried increasing the amount of training data, but the test accuracy only improved to 72%. Which additional approach is most likely to yield the largest improvement in test accuracy?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to CT-AI

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 ISTQB. “CT-AI” is a trademark of its owner, used for identification only.