AI and Machine Learning Fundamentals Practice Test

AI and Machine Learning Fundamentals Practice Test: all 15 questions with answers

Try each question first, then open “Show answer” to check the correct option and explanation.

1. Q1.Which statement best describes machine learning?

   • A.Software that only follows rules written by hand
   • B.Systems that learn patterns from data to make predictions or decisions
   • C.A database that stores answers to every question
   • D.A faster type of computer processor
   Show answer

   Correct answer: B. Systems that learn patterns from data to make predictions or decisions

   Machine learning systems learn patterns from example data instead of relying only on hand-written rules.

2. Q2.What does supervised learning need?

   • A.Labelled examples with inputs and the correct outputs
   • B.No data at all
   • C.Only rewards from an environment
   • D.Data without any labels
   Show answer

   Correct answer: A. Labelled examples with inputs and the correct outputs

   Supervised learning trains on labelled examples, such as emails marked spam or not spam.

3. Q3.A shop groups customers into segments without any labels. Which technique is this?

   • A.Regression
   • B.Classification
   • C.Clustering
   • D.Reinforcement learning
   Show answer

   Correct answer: C. Clustering

   Clustering is unsupervised learning: it finds groups in unlabelled data.

4. Q4.Predicting the selling price of a house is which kind of task?

   • A.Classification
   • B.Regression
   • C.Clustering
   • D.Dimensionality reduction
   Show answer

   Correct answer: B. Regression

   Regression predicts a continuous number, such as a price.

5. Q5.Deciding whether an email is spam or not spam is which kind of task?

   • A.Regression
   • B.Clustering
   • C.Classification
   • D.Translation
   Show answer

   Correct answer: C. Classification

   Classification assigns each input to one of a fixed set of categories.

6. Q6.A model scores 99% on training data but 60% on new data. What is the most likely problem?

   • A.Underfitting
   • B.Overfitting
   • C.Data leakage into the test set
   • D.The learning rate is zero
   Show answer

   Correct answer: B. Overfitting

   Overfitting means the model memorised the training data and does not generalise to unseen data.

7. Q7.Which change usually helps reduce overfitting?

   • A.Training for many more epochs on the same data
   • B.Removing the validation set
   • C.Adding regularisation such as dropout, or collecting more training data
   • D.Using a much larger model with the same small dataset
   Show answer

   Correct answer: C. Adding regularisation such as dropout, or collecting more training data

   Regularisation and more data both make it harder for the model to memorise noise.

8. Q8.Why do we keep a separate test set?

   • A.To make training faster
   • B.To estimate how the model performs on data it has never seen
   • C.To store extra training examples
   • D.To increase the training accuracy
   Show answer

   Correct answer: B. To estimate how the model performs on data it has never seen

   The test set is held back so it gives an honest estimate of real-world performance.

9. Q9.In reinforcement learning, how does an agent learn?

   • A.From labelled input-output pairs
   • B.By taking actions in an environment and receiving rewards or penalties
   • C.By clustering unlabelled data
   • D.By copying a fixed rule book
   Show answer

   Correct answer: B. By taking actions in an environment and receiving rewards or penalties

   Reinforcement learning optimises behaviour through trial and error guided by rewards.

10. Q10.How are the weights of a neural network usually updated during training?

    • A.They are chosen at random every step
    • B.By gradient descent, using gradients computed with backpropagation
    • C.They never change after initialisation
    • D.By a human editing them manually
    Show answer

    Correct answer: B. By gradient descent, using gradients computed with backpropagation

    Backpropagation computes how each weight affects the loss, and gradient descent moves the weights to reduce it.

11. Q11.What does 'deep' mean in deep learning?

    • A.The model is trained underground
    • B.The neural network has many layers
    • C.The data is very old
    • D.The model only works on images
    Show answer

    Correct answer: B. The neural network has many layers

    Deep learning uses neural networks with many stacked layers that learn increasingly abstract features.

12. Q12.Precision answers which question?

    • A.Of all actual positives, how many did we find?
    • B.Of everything we predicted as positive, how many were actually positive?
    • C.How fast is the model?
    • D.How many features does the model use?
    Show answer

    Correct answer: B. Of everything we predicted as positive, how many were actually positive?

    Precision = true positives / all predicted positives. It matters when false alarms are costly.

13. Q13.A cancer-screening model must miss as few real cases as possible. Which metric matters most?

    • A.Precision
    • B.Recall
    • C.Training speed
    • D.Model size
    Show answer

    Correct answer: B. Recall

    Recall = true positives / all actual positives. High recall means few real cases are missed.

14. Q14.In machine learning, what is a 'feature'?

    • A.A bug in the model
    • B.An input variable the model uses to make a prediction
    • C.The final prediction
    • D.The name of the dataset
    Show answer

    Correct answer: B. An input variable the model uses to make a prediction

    Features are the measurable inputs, such as square footage or number of rooms for a house-price model.

15. Q15.What makes generative AI different from a classic classifier?

    • A.It can only answer yes or no
    • B.It creates new content such as text, images or code
    • C.It never uses neural networks
    • D.It does not need any training data
    Show answer

    Correct answer: B. It creates new content such as text, images or code

    Generative models produce new content that follows the patterns they learned, rather than only assigning a label.