Bayesian Estimation (BE)Bayesian Parameter Estimation ：贝叶斯估计（BE）的贝叶斯参数估计.ppt

Pattern Classification Part 2 Dr. Ding Yuxin Pattern Classification Part 2 * Bayesian Estimation (BE) Bayesian Parameter Estimation: Gaussian Case Bayesian Parameter Estimation: General Estimation Chapter 3Maximum-Likelihood and Bayesian Parameter Estimation (3,4,5) Dr. Ding Yuxin Pattern Classification Part 2 * In MLE ? was supposed fix In BE ? is a random variable The computation of posterior probabilities P(?i | x) lies at the heart of Bayesian classification Goal: compute P(?i | x, D) Given the sample D, Bayes formula can be written 3.3 Bayesian Estimation Dr. Ding Yuxin Pattern Classification Part 2 * To demonstrate the preceding equation, use: Dr. Ding Yuxin Pattern Classification Part 2 * Parameter Distribution is unknown, we assume it has a known parametric form , and value of parameter is unknown Know prior density Training data convert to a posterior density Our path: Dr. Ding Yuxin Pattern Classification Part 2 * If peaks very sharply about parameter and is smooth, and the tails of the integral are not important, then Dr. Ding Yuxin Pattern Classification Part 2 * 3.4 Bayesian Parameter Estimation: Gaussian Case Goal: Estimate ? using the a-posteriori density P(? | D) The univariate case: P(? | D) ? is the only unknown parameter (?0 and ?0 are known!) Dr. Ding Yuxin Pattern Classification Part 2 * Reproducing density Identifying (1) and (2) yields: 4 Dr. Ding Yuxin Pattern Classification Part 2 * Understanding represents our best guess for after observing n samples measures our uncertainty about this guess Add samples to decrease the uncertainty Bayse Learning: as n increase, p(? | D) becomes more and more sharply peaked, approaching a Dirac delta function as n approaches infinity Dr. Ding Yuxin Pattern Classification Part 2 * Dr. Ding Yuxin Pattern Classification Part 2 * The univariate case P(x | D) P(? | D) computed as a