In the Bias-Variance decomposition of prediction error, what does 'Bias' reflect?
Bias measures the systematic error caused by simplifying assumptions (e.g., assuming linear relationships for non-linear data).
Which techniques are effective for mitigating Overfitting in complex machine learning models?
Regularization, tree pruning, dropout, and increasing dataset size are all standard, proven defenses against overfitting.
What distinguishes L1 Regularization (Lasso) from L2 Regularization (Ridge)?
L1 (Lasso) adds lambda * sum(|w_i|), driving less important coefficients to zero, whereas L2 (Ridge) adds lambda * $sum(w_i^2)$, shrinking weights toward zero without setting them exactly to zero.
In evaluating binary classifiers, what formula calculates 'Precision'?
Precision measures the proportion of predicted positive instances that are truly positive: Precision = TP / (TP + FP).
In classification diagnostics, what is 'Recall' (also known as Sensitivity or True Positive Rate)?
Recall measures the proportion of actual positive cases successfully identified: Recall = TP / (TP + FN).