Machine Learning

Free CFA Level II lesson in Quantitative Methods. 12 min read, ~1,827 words.

Bias-variance tradeoff: complex models overfit (high variance), simple models underfit (high bias). Diagnose from the training/test gap. LASSO (L1) sets coefficients to exactly zero = automatic feature selection. Ridge (L2) shrinks but never zeroes. Bagging/random forests reduce variance. Boosting reduces bias. Do not reverse these. K-means requires pre-specifying k: the...