Explain core ML concepts and metrics

You are interviewing for a **Data Scientist** role. Answer the following ML fundamentals questions clearly and concisely. ### Concepts 1. Explain the **bias–variance tradeoff**. How does it relate to **overfitting vs. underfitting**? 2. What are common forms of **regularization** (e.g., L1/L2, early stopping)? What problem does regularization solve? ### Imbalanced classification 3. For **imbalanced datasets**, which evaluation metrics do you prefer and why? Compare **accuracy, precision, recall, F1, PR-AUC, ROC-AUC**. 4. Define **precision** and **recall** and provide their formulas. When would you optimize for one over the other? ### Logistic regression / probabilities 5. In **logistic regression**, what is the model’s **raw output** before converting to a probability? How is it mapped to a probability? ### ROC-AUC interpretation 6. What does it mean if **ROC-AUC = 0.8**? Provide an intuitive interpretation and at least one caveat. ### Models 7. What are **ensemble models** and why do they often outperform a single model? 8. For **tree-based models** (decision trees / random forests / gradient boosting), name key **hyperparameters** and describe how they affect bias/variance. ### Output activations 9. Compare **sigmoid vs. softmax**: when do you use each, and how do their outputs differ?