Online Learning

Online learning is useful when data changes rapidly or arrives as a stream and waiting for periodic full retraining would be too slow. Production use requires controls for data quality, drift, catastrophic changes, and rollback if new updates reduce model performance.