Exercises: ML Fundamentals

Concept Drills

• Explain bias and variance with a drawing.
• Explain train, validation, and test splits.
• Compare precision, recall, F1, ROC-AUC, and PR-AUC.
• Explain why data leakage invalidates evaluation.

Coding Drills

• Implement mean, variance, and standard deviation from scratch.
• Implement linear regression with gradient descent.
• Train logistic regression on a classification dataset.
• Compare decision tree, random forest, and gradient boosting.
• Build a reusable evaluation function.

Failure Drills

• Create an overfitting example.
• Create an underfitting example.
• Create a leakage example.
• Train on imbalanced data and compare metrics.

Output

• Notebook or script for each drill.
• One evaluation report.
• One README section explaining trade-offs.