AdaBoost builds strong classifiers from weak decision

Short summary

AdaBoost turns weak learners (decision stumps barely better than random guessing) into strong classifiers through iterative reweighting. After each round, misclassified points gain weight, forcing the next stump to focus on hard examples. Final predictions combine weighted votes from all stumps, with alpha controlling each stump's influence based on training error.

• •Weak learners (decision stumps) achieve >50% accuracy by reweighting misclassified points with each iteration
• •Each stump's confidence (alpha) = 0.5 × ln((1-error)/error); near-perfect stumps dominate the vote
• •Ensemble accuracy climbs from ~55% to 85%+ over 25 rounds despite individual stumps never improving

Generated with AI, which can make mistakes.