Classification

Boosting.md

Stump weights (𝐰̂) and data point weights (𝛼) are two different concepts. Stump weights (𝐰̂) tell you how important each stump is while making predictions with the entire boosted ensemble. Data point weights (𝛼) tell you how important each data point is while training a decision stump.

Greedy Decision Tree Algorithm.md

Selecting the best feature to split on Make a decision tree for each feature M Step 1: Start with an empty tree Step 2: Select the best feature to split on Select feature with lowest classification error To compute the error for each feature M: * for each stump node divide the number of incorrect classifications by the total number of classifications Once you have selected best feature for each split in tree: Step 3: if no more features to split on, make predictions Step 4: otherwise recurse back to step 2

Stopping Criteria Stop splitting once all decision stumps are classified OR out of features to split on