手法を比較

	ロバストブースティング	正則化ブースティング	ロバスト勾配ブースティング
分野	機械学習	機械学習	機械学習
系統	Machine learning	Machine learning	Machine learning
提唱年	1999–2001	2001–2016	2001
提唱者	Freund, Y.; Mason, L. et al.	Friedman, J. H.; extended by Chen & Guestrin	Friedman, J. H. (with Huber loss from Huber, P. J.)
種類	Ensemble (robust sequential boosting)	Regularized ensemble (boosting with shrinkage/penalty)	Ensemble (boosted trees with robust loss)
原典	Freund, Y. (2001). An adaptive version of the boost by majority algorithm. Machine Learning, 43(3), 293–318.	Friedman, J. H. (2001). Greedy function approximation: A gradient boosting machine. Annals of Statistics, 29(5), 1189–1232.	Friedman, J. H. (2001). Greedy function approximation: A gradient boosting machine. Annals of Statistics, 29(5), 1189–1232.
別名	noise-tolerant boosting, robust AdaBoost, boosting with robust losses, outlier-resistant boosting	shrinkage boosting, penalized boosting, regularized gradient boosting, L1/L2 boosting	gradient boosting with Huber loss, robust GBM, outlier-robust boosting, robust gradient-boosted trees
関連	6	5	6
概要	Robust Boosting modifies standard boosting algorithms — such as AdaBoost or gradient boosting — by replacing the default exponential or squared loss with robust loss functions (e.g., Huber, logistic, or truncated losses) or by incorporating noise-tolerance mechanisms, so that the ensemble remains accurate even when training data contain outliers, label noise, or heavy-tailed errors.	Regularized boosting extends gradient boosting by adding explicit controls — shrinkage (learning rate), L1/L2 weight penalties, subsampling, and tree-complexity limits — to the objective function and the update rule. These constraints reduce overfitting, stabilise the model on noisy or small datasets, and are the core reason why systems such as XGBoost and LightGBM consistently outperform vanilla boosting on real-world tabular benchmarks.	Robust Gradient Boosting is gradient boosting trained with outlier-resistant loss functions — most commonly the Huber loss or quantile (pinball) loss — instead of squared-error loss. Proposed in Friedman's seminal 2001 paper, this variant produces predictions far less distorted by extreme values or contaminated labels, while retaining the full predictive power of gradient-boosted trees.