Supervised Machine Learning
Derive the workhorse supervised methods rather than merely calling them: least squares, logistic regression, shrinkage penalties, and tree ensembles.
Sign in to take quizzes, earn XP, and unlock stages as you reach 90% mastery.
Least Squares from the Derivative
30 min · 120 XPMinimising the residual sum of squares in closed form, what the slope formula means, and why R-squared cannot compare models of different sizes.
Open lesson →Sign in to take the 2-question quiz.
Classification and the Log-Odds
30 min · 120 XPWhy a linear probability model is impossible, what the logistic coefficients actually mean, and how maximum likelihood fits them.
Open lesson →Sign in to take the 2-question quiz.
Shrinkage: Ridge and Lasso
25 min · 120 XPTrading a little bias for a large variance reduction, and the geometric reason the L1 penalty produces exact zeros where L2 does not.
Open lesson →Sign in to take the 2-question quiz.
Trees, Bagging, and Random Forests
30 min · 140 XPRecursive binary splitting, why Gini beats accuracy as a criterion, and how deliberately handicapping each tree improves the ensemble.
Open lesson →Sign in to take the 2-question quiz.