CAIC Machine Learning & Data Science 3 — Questions and Answers
Question 1: Which ensemble method trains multiple models in parallel on random subsets of the training data and aggregates their predictions?
- Boosting
- Stacking
- Bagging (Correct answer)
- Bayesian averaging
Correct answer: Bagging
Bagging (Bootstrap Aggregating) trains independent learners on bootstrapped samples and combines them via voting or averaging to reduce variance.
Question 2: What does the ROC curve plot?
- Precision vs. Recall at different thresholds
- True Positive Rate vs. False Positive Rate at different thresholds (Correct answer)
- Training loss vs. Validation loss over epochs
- Model accuracy vs. Dataset size
Correct answer: True Positive Rate vs. False Positive Rate at different thresholds
The ROC curve shows the tradeoff between sensitivity (TPR) and specificity (1-FPR) across all classification thresholds.
Question 3: Which regularization technique randomly sets a fraction of neuron activations to zero during each training step?
- L1 regularization
- Weight decay
- Dropout (Correct answer)
- Early stopping
Correct answer: Dropout
Dropout randomly deactivates neurons during training, forcing the network to learn redundant representations and reducing co-adaptation.
Question 4: In a confusion matrix for binary classification, which formula correctly computes Precision?
- TP / (TP + FN)
- TP / (TP + FP) (Correct answer)
- TN / (TN + FP)
- (TP + TN) / (TP + TN + FP + FN)
Correct answer: TP / (TP + FP)
Precision measures what fraction of predicted positives are actually positive: TP divided by all predicted positives (TP + FP).
Question 5: What is the key difference between supervised and unsupervised learning?
- Supervised learning uses neural networks; unsupervised uses decision trees
- Supervised learning requires labeled training data; unsupervised discovers patterns without labels (Correct answer)
- Supervised learning is faster to train than unsupervised learning
- Supervised learning works only on tabular data; unsupervised works on images
Correct answer: Supervised learning requires labeled training data; unsupervised discovers patterns without labels
Supervised learning learns a mapping from inputs to known labels, while unsupervised learning finds hidden structure in unlabeled data.
Question 6: Which metric measures the average magnitude of errors in the same units as the target variable, without squaring them?
- Mean Squared Error (MSE)
- Root Mean Squared Error (RMSE)
- Mean Absolute Error (MAE) (Correct answer)
- R-squared
Correct answer: Mean Absolute Error (MAE)
MAE computes the mean of absolute differences between predictions and actuals, keeping error in the original unit scale without amplifying outliers.
Question 7: What is 'feature engineering' in the context of machine learning?
- Selecting the best algorithm for a given dataset
- Transforming raw data into informative input representations for a model (Correct answer)
- Tuning hyperparameters using grid search
- Splitting data into training and test sets
Correct answer: Transforming raw data into informative input representations for a model
Feature engineering involves creating, transforming, or selecting variables from raw data to improve model performance.
CAIC Machine Learning & Data Science 3
Which ensemble method trains multiple models in parallel on random subsets of the training data and aggregates their predictions?