01
True or false
1 point
True or false: Unsupervised learning requires answer labels to be supplied for every example.
Free Practice Quiz Question List
4 Machine Learning Basics Online Quiz Questions
Use this free practice quiz with 20 questions to review 4 Machine Learning Basics, test your knowledge, and prepare for your next test or exam.
20 questions
02
Choose one
1 point
A model uses an apartment's size, location, and age to predict its sale price. Which kind of task is this?
- A
Classification, because each apartment belongs to a location category
- B
Regression, because the predicted sale price is a numerical value
- C
Clustering, because the model uses several input features
- D
Dimensionality reduction, because apartment attributes are used as inputs
03
Written response
1 point
What is the machine-learning term for a model's ability to perform well on new examples?
04
Fill in the blank
1 point
In a supervised email example, the email's words are called its , while the desired answer, such as “spam,” is its .
05
True or false
1 point
True or false: Repeatedly changing a model based on test-set results can make the final performance estimate less trustworthy.
- A
True
- B
False
06
Written response
1 point
What is the term for information that would not genuinely be available at prediction time sneaking into model training or evaluation?
07
Choose one
1 point
Which statement correctly distinguishes a model from an algorithm in machine learning?
- A
The model is the procedure that adjusts parameters; the algorithm is the learned function.
- B
The model is the dataset; the algorithm is the label assigned to each example.
- C
The model is the learned function; the algorithm is the procedure that adjusts it.
- D
The model is the loss measure; the algorithm is the prediction error.
08
Choose one
1 point
A team is building a model to predict future demand from historical data. Which train-test split best respects the order of prediction?
- A
Randomly mix all dates across training and test sets to give each set a similar time range.
- B
Train on earlier dates and test on later dates.
- C
Train on later dates and test on earlier dates.
- D
Use the same examples for training and testing so performance can be compared directly.
09
Choose all
1 point
Select all statements that correctly match a classification measure to the cost it helps address.
- A
Precision is useful when false alarms are costly.
- B
Recall is the fraction of predicted positives that truly are positive.
- C
Recall is useful when missing positive cases is costly.
- D
Precision is the fraction of all examples that the model classifies correctly.
10
Fill in the blank
1 point
When development data is divided into parts, each held-out part serves as the in turn; repeating this process is called .
11
Choose all
1 point
Select all statements that correctly describe the goals or mechanics of model training.
- A
Training adjusts model parameters to reduce a numerical measure of prediction error.
- B
The goal of training is to memorize the training examples, even if performance on new examples suffers.
- C
A model that performs well on new examples demonstrates generalization.
- D
The algorithm is the learned function, while the model is the procedure that adjusts it.
12
Open ended
1 point
A team is evaluating a classifier on a dataset where one class is much more common than the other. Describe a suitable simple baseline and explain why comparing accuracy alone with that baseline may be insufficient.
13
Choose one
1 point
Which is an example of a simple baseline for a classification task?
- A
Train a complex model and repeatedly tune it using test-set results.
- B
Always predict the most frequent class.
- C
Use the validation set as the only measure of performance on unseen data.
- D
Fit model parameters using the test set.
14
Choose one
1 point
A team trains a model on past transactions labeled as fraudulent or legitimate so it can predict whether new transactions are fraudulent. Which learning approach is it using?
- A
Unsupervised learning
- B
Supervised learning
- C
Reinforcement learning
- D
Dimensionality reduction
15
Choose one
1 point
A team wants to know whether its classifier performs better than a simple reference. Which option is an appropriate classification baseline?
- A
A model that uses every available feature
- B
A model evaluated only on its training examples
- C
A simple reference prediction, such as always choosing the most frequent class
- D
A model whose predictions are always correct
16
Choose one
1 point
During model development, a team compares different feature choices before making a final evaluation. Which dataset subset should it use for this comparison?
- A
Validation set
- B
Training set
- C
Test set
- D
Unlabeled set
17
Choose one
1 point
A model performs very well on its training examples but substantially worse on new examples. Which problem does this pattern indicate?
- A
The model has learned to generalize well
- B
The model is using a baseline
- C
The model is performing dimensionality reduction
- D
The model has overfit the training examples
18
True or false
1 point
True or false: R² compares a model’s predictions with a simple mean-based baseline, and it should be interpreted alongside error measures and the task context.
- A
True
- B
False
19
Written response
1 point
What term names the learning procedure that adjusts a model’s parameters?
20
Written response
1 point
A project has a small set of labeled examples and a much larger set of unlabeled examples. What learning approach uses both sets?