Machine Learning Concepts Quiz

Grade 11 · Technology · Quiz

Eight questions covering supervised and unsupervised learning, neural networks, and model evaluation.

Questions

1. Which of the following is an example of an unsupervised learning task?

   1. Predicting house prices from square footage
   2. Classifying email as spam or not spam
   3. Grouping customers by purchasing behavior
   4. Recognizing handwritten digits
   Show answer

   Answer: Grouping customers by purchasing behavior

   Clustering customers by behavior is unsupervised — there are no predefined labels; the algorithm discovers natural groupings in the data.

2. A model with very high training accuracy but low test accuracy is most likely suffering from underfitting.

   Show answer

   Answer: False

   High training accuracy combined with low test accuracy is the hallmark of overfitting, not underfitting; underfitting yields poor accuracy on both sets.

3. In a neural network, the ________ algorithm calculates gradients layer by layer from the output back to the input.

   Show answer

   Answer: backpropagation

   Backpropagation applies the chain rule of calculus to propagate the loss gradient from the output layer backward through every layer, enabling parameter updates.

4. Which technique helps prevent overfitting by randomly deactivating a fraction of neurons during each training step?

   1. Batch normalization
   2. Dropout
   3. Weight initialization
   4. Learning rate scheduling
   Show answer

   Answer: Dropout

   Dropout randomly sets neuron outputs to zero during training, preventing co-adaptation and acting as a regularizer that reduces overfitting.

5. Principal Component Analysis (PCA) increases the dimensionality of a dataset.

   Show answer

   Answer: False

   PCA reduces dimensionality by projecting data onto a smaller set of principal components that capture the most variance, making it easier to visualize and process.

6. A decision tree classifies an input by routing it through a series of ________ tests on feature values.

   Show answer

   Answer: binary

   A decision tree routes each example through a sequence of binary (yes/no) tests on feature values until reaching a leaf node that predicts the class.

7. Which evaluation metric is most appropriate when false negatives are far more costly than false positives (e.g., cancer screening)?

   1. Precision
   2. Recall
   3. Accuracy
   4. Specificity
   Show answer

   Answer: Recall

   Recall measures how many actual positives the model catches; in cancer screening, missing a real case (false negative) is far worse than a false alarm, making recall the priority metric.

8. In gradient descent, the learning rate controls how large a step is taken in the direction of the gradient on each update.

   Show answer

   Answer: True

   The learning rate is a hyperparameter that scales the gradient update; too large causes oscillation or divergence, too small causes very slow convergence.

Get a personalized course on Machine Learning Concepts built around your child.