Embedded Files
A decision stump is a one-level decision tree with a root node connected directly to leaf nodes. It evaluates just one feature—often using a threshold for continuous values—to classify or regress data into categories like "above/below threshold".
Decision Stump Implementation
Decision Stump Implementation
I implemented AdaBoost from scratch with entropy-based decision stumps. The training logs show classic boosting behavior: early stumps captured coarse structure, later stumps specialized on hard boundary regions around feature values 4.5–5.5. Weak learners with error ≥ 0.5 were automatically rejected, demonstrating robustness. After 20 rounds, the ensemble achieved zero training error by combining multiple piecewise decision boundaries
Page updated
Google Sites
Report abuse