方法对比

	经验贝叶斯	Bayesian Regression	岭回归（Ridge Regression）
领域	贝叶斯	贝叶斯	机器学习
方法族	Bayesian methods	Bayesian methods	Machine learning
起源年份	—	—	1970
提出者	Herbert Robbins (1956); Bradley Efron & Carl Morris (1973)	—	Hoerl, A.E. & Kennard, R.W.
类型	Empirical Bayes estimator	Bayesian linear model	L2-regularized linear regression
开创性文献	Robbins, H. (1956). An empirical Bayes approach to statistics. In J. Neyman (Ed.), Proceedings of the Third Berkeley Symposium on Mathematical Statistics and Probability, Vol. 1 (pp. 157–164). University of California Press.	Gelman, A., Carlin, J. B., Stern, H. S., Dunson, D. B., Vehtari, A. & Rubin, D. B. (2013). Bayesian Data Analysis (3rd ed.). CRC Press. ISBN: 978-1439840955	Hoerl, A.E. & Kennard, R.W. (1970). Ridge Regression: Biased Estimation for Nonorthogonal Problems. Technometrics, 12(1), 55–67.
别名	EB, empirical Bayes estimation, marginal likelihood estimation, James-Stein shrinkage	bayesian linear regression, probabilistic regression, bayesian regresyon	Ridge Regresyonu, ridge regresyonu, L2-regularized regression, Tikhonov regularization
相关	4	2	4
摘要	Empirical Bayes (EB) is an estimation strategy, introduced by Herbert Robbins in 1956 and developed into practical shrinkage estimators by Bradley Efron and Carl Morris in 1973, in which the hyperparameters of the prior distribution are estimated from the observed data via the marginal likelihood rather than specified in advance. The resulting posterior retains a Bayesian structure but substitutes data-driven hyperparameters for subjective ones, bridging frequentist shrinkage and full Bayesian inference.	Bayesian regression is a probabilistic version of linear regression that treats the model parameters as uncertain quantities. Instead of returning a single best-fit estimate, it combines prior knowledge with the observed data to produce a full posterior probability distribution for each parameter, from which credible intervals and predictions are read off.	Ridge Regression is an L2-regularized linear regression method, introduced by Arthur Hoerl and Robert Kennard in 1970, that reduces multicollinearity by adding a penalty on the size of the coefficients. It shrinks coefficients toward zero without setting any of them exactly to zero, producing more stable estimates when predictors are highly correlated.