AdaBoost
Adaptive boosting ensemble that focuses on misclassified examples
AdaBoost sequentially trains weak classifiers (typically shallow trees), with each iteration re-weighting examples that were previously misclassified. This produces a strong classifier from simple base learners.
When to use:
- Binary classification with noisy data where boosting helps focus on hard examples
- When a simple, explainable ensemble is preferred over deep gradient boosting
- Datasets of moderate size where training speed matters
Input: Tabular data with the feature columns defined during training Output: Predicted class label and class probabilities
Model Settings (set during training, used at inference)
N Estimators (default: 50) Maximum number of estimators (boosting rounds). Training stops early if perfect fit is achieved.
Learning Rate (default: 1.0) Shrinks each estimator's contribution. Lower values require more estimators but improve generalization.
Algorithm (default: SAMME.R)
Boosting algorithm. SAMME.R uses probability estimates; SAMME uses discrete class labels.
Base Estimator Max Depth (default: 1) Depth of the decision tree stump base learner. Depth 1 (stumps) is the classic choice.
Inference Settings
No dedicated inference-time settings. The weighted combination of all estimators produces the final prediction.