DSE Knowledge 4 — Questions and Answers
Question 1: What distinguishes supervised learning from unsupervised learning?
- Supervised learning uses more data
- Supervised learning trains on labeled input-output pairs; unsupervised learning finds structure in unlabeled data (Correct answer)
- Unsupervised learning requires more computing power
- Supervised learning only works with numerical data
Correct answer: Supervised learning trains on labeled input-output pairs; unsupervised learning finds structure in unlabeled data
Supervised learning maps inputs to known outputs using labeled examples, while unsupervised learning discovers hidden patterns in data without labels.
Question 2: Which metric is most appropriate for evaluating a classifier on a heavily imbalanced dataset?
- Accuracy
- F1-Score or AUC-ROC (Correct answer)
- Mean Squared Error
- R-squared
Correct answer: F1-Score or AUC-ROC
F1-Score and AUC-ROC account for class imbalance, whereas accuracy can be misleadingly high when the majority class dominates.
Question 3: In gradient descent, what does the learning rate control?
- The number of training epochs
- The step size taken in the direction of the negative gradient at each update (Correct answer)
- The proportion of data used per batch
- The depth of a neural network
Correct answer: The step size taken in the direction of the negative gradient at each update
The learning rate scales the gradient update step; too large causes divergence, too small causes slow convergence.
Question 4: What type of relationship does Pearson correlation measure?
- Any monotonic relationship between two variables
- Linear relationship between two continuous variables (Correct answer)
- Causal relationship between variables
- Rank-based association between ordinal variables
Correct answer: Linear relationship between two continuous variables
Pearson correlation measures the strength and direction of the linear association between two continuous variables, ranging from -1 to +1.
Question 5: Which of the following is a key assumption of linear regression?
- Residuals are uniformly distributed
- There is a linear relationship between predictors and the response variable (Correct answer)
- All predictor variables must be binary
- The response variable must be bounded between 0 and 1
Correct answer: There is a linear relationship between predictors and the response variable
Linear regression assumes the expected value of the outcome is a linear combination of the predictor variables.
Question 6: What is the role of a validation set during model training?
- It is used to compute the final reported performance of the model
- It provides unbiased data for tuning hyperparameters without contaminating the test set (Correct answer)
- It replaces the training set in small datasets
- It is used only for neural networks, not traditional ML
Correct answer: It provides unbiased data for tuning hyperparameters without contaminating the test set
The validation set allows iterative hyperparameter tuning while keeping the test set pristine for final unbiased evaluation.
Question 7: Which ensemble method trains multiple models sequentially, with each model focusing on examples the previous one misclassified?
- Bagging
- Random Forests
- Boosting (Correct answer)
- Stacking
Correct answer: Boosting
Boosting trains models sequentially, re-weighting misclassified examples so subsequent learners correct earlier mistakes.
DSE Knowledge 4
What distinguishes supervised learning from unsupervised learning?