Hands-On Exercise: Implement a Simple Perceptron CT-AI Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   In a binary classification task using a perceptron, what does the output of the step activation function represent?

   • AThe distance of the input from the decision boundary.
   • BThe class label (e.g., 0 or 1) predicted for the input.
   • CThe confidence of the prediction.
   • DThe probability that the input belongs to the positive class.
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   A tester is writing unit tests for a perceptron implementation. Which test case is most important to verify the perceptron learning rule correctly updates weights?

   • AVerify the activation function outputs values between 0 and 1.
   • BFeed a single sample with a known error and assert the weights change by the expected amount.
   • CCheck that the bias remains constant during training.
   • DFeed a large batch of samples and assert the loss decreases monotonically.
   Reveal Solution Discussion Select an answer first
4. 8foundation · easy

   What does it mean for a perceptron to converge during training?

   • AThe weights become zero.
   • BThe training data is split into training and validation sets.
   • CThe learning rate is reduced to zero.
   • DThe weights stop changing because the perceptron correctly classifies all training samples.
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A team is training a perceptron with learning rate η = 0.5. After several epochs, the weights are oscillating and the model does not converge on a linearly separable dataset. What is the most appropriate adjustment?

   • ADecrease the learning rate to 0.1
   • BInitialize the weights to zero
   • CIncrease the learning rate to 1.0
   • DAdd more training epochs
   Reveal Solution Discussion Select an answer first
6. 10expert · hard

   A team is using a perceptron to classify whether a loan application is approved (1) or denied (0) based on income and credit score. The data is linearly separable, but the model's accuracy on a held-out test set is only 70%. What is the most likely explanation?

   • AThe perceptron requires more training epochs to converge to a better solution.
   • BThe test set contains samples that are not linearly separable from the training data.
   • CThe perceptron has overfit the training data because the learning rate is too high.
   • DThe decision boundary found by the perceptron is one of many possible separating boundaries, and it may not generalize well to unseen data.
   Reveal Solution Discussion Select an answer first