Ensemble method

Machine learning technique that combines multiple models (votes or averages) to improve accuracy and stability, including bagging and random forests.

An ensemble method trains many models on the same problem and aggregates their outputs, often by majority vote or averaging. Bagging and random forests (Breiman, article) are classic examples that reduce variance.

The grand-prize BellKor’s Pragmatic Chaos team blended multiple predictors to win the Netflix Prize, even though Netflix later favored simpler factorization and RBM components for production. Ensembles also power real-time pose estimation in the Kinect depth pipeline.