方法对比

	在线梯度提升	Boosting	在线学习
领域	机器学习	机器学习	机器学习
方法族	Machine learning	Machine learning	Machine learning
起源年份	2011–2015	1990–1997	1958–2000s
提出者	Grubb, A. & Bagnell, J. A.; Beygelzimer, A. et al.	Schapire, R. E.; Freund, Y.	Rosenblatt, F.; Littlestone, N.; Shalev-Shwartz, S. (key contributors)
类型	Online ensemble (sequential boosting on streaming data)	Sequential ensemble (iterative reweighting)	Learning paradigm (sequential model update)
开创性文献	Grubb, A. & Bagnell, J. A. (2011). Generalized Boosting Algorithms for Convex Optimization. Proceedings of the 28th International Conference on Machine Learning (ICML 2011), 1209–1216.	Freund, Y. & Schapire, R. E. (1997). A decision-theoretic generalization of on-line learning and an application to boosting. Journal of Computer and System Sciences, 55(1), 119–139.	Shalev-Shwartz, S. (2011). Online Learning and Online Convex Optimization. Foundations and Trends in Machine Learning, 4(2), 107–194.
别名	OGB, streaming gradient boosting, incremental gradient boosting, online boosting with gradient descent	AdaBoost, gradient boosting, iterative reweighting ensemble, sequential ensemble	incremental learning, sequential learning, streaming learning, online machine learning
相关	6	6	6
摘要	Online Gradient Boosting adapts the gradient boosting framework for streaming settings where data arrives one sample at a time rather than as a fixed batch. At each step the model computes a pseudo-residual for the incoming observation and updates a weak learner in place, growing an additive ensemble without storing or revisiting past data. This makes it suitable for real-time prediction and large-scale streaming pipelines where retraining from scratch is infeasible.	Boosting is a sequential ensemble technique that converts many simple, barely-better-than-chance learners into a single highly accurate model by repeatedly focusing training on the examples that previous learners got wrong, then combining all learners with weights proportional to their individual accuracy.	Online learning is a machine learning paradigm in which a model is updated incrementally as each new data point arrives, rather than being trained once on a fixed dataset. It is essential when data streams continuously, storage is limited, or the underlying distribution shifts over time. Theoretical performance is measured by cumulative regret relative to the best fixed predictor in hindsight.