01
Choose one
1 point
A model performs well on its training data but much worse on separate validation data. Which issue is this pattern most likely to indicate?
Free Practice Quiz Question List
8 Overfitting and Generalization Online Quiz Questions
Use this free practice quiz with 20 questions to review 8 Overfitting and Generalization, test your knowledge, and prepare for your next test or exam.
20 questions
02
Written response
1 point
A model performs poorly on both its training data and validation data. What modeling problem does this pattern most likely indicate?
03
True or false
1 point
Generalization describes a model's ability to make accurate predictions on previously unseen data.
- A
True
- B
False
04
Choose one
1 point
A very high-degree polynomial passes through every training point, including unusual sales, but predicts prices poorly for other houses. What is the best explanation?
- A
It is underfitting because it cannot match the training points.
- B
It is guaranteed to generalize well because its training error is low.
- C
It may be overfitting by capturing noise or accidental details in the training data.
- D
It is experiencing data leakage because the polynomial has many terms.
05
Choose all
1 point
A model is overfitting. Which actions could help reduce its effective complexity or prevent it from continuing to fit training-data quirks? Select all that apply.
- A
Limit a decision tree's depth.
- B
Use regularization to discourage overly complex parameter values.
- C
Increase the model's flexibility so it can fit every training example more closely.
- D
Use early stopping when validation performance stops improving.
06
Fill in the blank
1 point
Use to compare models or tune settings, and keep the test set for final evaluation.
07
Written response
1 point
What problem can occur if you perform preprocessing or feature selection before splitting a dataset into training, validation, and test sets? Enter the established term.
08
Choose all
1 point
Which practices help prevent data leakage during preprocessing? Select all that apply.
- A
Split the data before preprocessing or feature selection.
- B
Fit scaling or imputation on training data, then apply the learned transformation to validation and test data.
- C
Fit preprocessing steps on the full dataset before creating the training and validation splits.
- D
Use pipelines to keep preprocessing steps within the correct data split.
09
True or false
1 point
In k-fold cross-validation, the model trains on all but one fold and validates on the remaining fold, repeating until every fold has served as validation data.
- A
True
- B
False
10
Choose one
1 point
A model appears to underfit because strong regularization is preventing it from learning useful patterns. Which change is most appropriate?
- A
Increase regularization further.
- B
Reduce excessive regularization.
- C
Limit decision-tree depth even more.
- D
Stop training earlier.
11
Fill in the blank
1 point
In k-fold cross-validation, each fold serves once as the while the model trains on all the other folds.
12
Choose one
1 point
Which tools can help reveal patterns in model performance as training or model settings vary?
- A
A final test score alone, without comparing performance patterns.
- B
A list of training examples in their original order.
- C
Learning curves and validation curves.
- D
The number of parameters, considered without any performance measurements.
13
True or false
1 point
When comparing model performance, lower mean squared error is better, while higher accuracy is better.
- A
True
- B
False
14
Open ended
1 point
Explain how validation and test data should be used during model development, and why repeatedly choosing settings based on test performance can undermine the final evaluation.
15
Written response
1 point
What machine-learning term describes the ability to make accurate predictions on previously unseen data?
16
Choose one
1 point
A deep decision tree performs well on training data but much worse on validation data. Which change most directly addresses the likely problem?
- A
Increase the tree depth so it can fit the training data more closely.
- B
Limit the tree depth to reduce the model's effective complexity.
- C
Choose settings by repeatedly checking the test score.
- D
Fit preprocessing steps using the validation data.
17
Choose one
1 point
A model's training and validation performance are both poor. The model is known to be too simple, and training has converged. Which change is most appropriate?
- A
Use a simpler model and increase regularization.
- B
Reduce the amount of training data.
- C
Try a more suitable model or add useful features.
- D
Choose model settings using only the test score.
18
Written response
1 point
Name one type of curve that can help reveal overfitting or underfitting patterns.
19
Choose one
1 point
A team discovers that its training examples do not resemble the situations in which its predictions will be used. What should it prioritize to support generalization?
- A
Check that the training examples are representative of the situations where predictions will be used.
- B
Use only examples that are easy for the current model to predict.
- C
Select training examples based on the test score.
- D
Make the model more complex regardless of the data.
20
Choose one
1 point
When dividing data for model development, which practice correctly distinguishes validation data from the test set?
- A
Use the test set to tune settings, then use validation data for the final evaluation.
- B
Use validation data to compare models or tune settings, and reserve the test set for final evaluation.
- C
Use validation data to fit the model and training data only for the final evaluation.
- D
Use training data for tuning and repeatedly check the test set after every change.