http://wiki.christophchamp.com/index.php?action=history&feed=atom&title=Maximum_a_posteriori 2026-04-30T06:43:26Z Revision history for this page on the wiki MediaWiki 1.26.2 http://wiki.christophchamp.com/index.php?title=Maximum_a_posteriori&diff=1534&oldid=prev 2005-12-29T23:05:24Z

<p>Started article</p> <p><b>New page</b></p><div>In [[statistics]], the method of '''maximum a posteriori''' (MAP, posterior mode, or '''maximum posterior probability''') [[estimation theory|estimation]] can be used to obtain a [[point estimation|point estimate]] of an unobserved quantity on the basis of empirical data. It is closely related to Ronald Fisher's method of [[maximum likelihood]] (ML), but employs an augmented optimization objective which incorporates a [[prior distribution]] over the quantity one wants to estimate. MAP estimation can therefore be seen as a [[regularization (mathematics)|regularization]] of ML estimation.<br /> <br /> MAP estimates can be computed in several ways:<br /> # Analytically, when the mode(s) of the posterior distribution can be given in [[closed-form solution|closed form]]. This is the case when [[conjugate prior]]s are used.<br /> # Via [[numerical analysis|numerical]] [[optimization (mathematics)|optimization]] such as the [[conjugate gradient method]] or [[Newton's method in optimization|Newton's method]]. This usually requires first or second [[derivative]]s, which have to be evaluated analytically or numerically.<br /> # Via a modification of an [[expectation-maximization algorithm]]. This does not require derivatives of the posterior density.<br /> <br /> While MAP estimation shares the use of a prior distribution with Bayesian statistics, it is not generally seen as a Bayesian method. This is because MAP estimates are point estimates, whereas Bayesian methods are characterized by the use of distributions to summarize data and draw inferences. Bayesian methods tend to report the posterior [[mean]] or [[median]] together with [[posterior interval]]s, rather than the posterior mode. This is especially so when the posterior distribution does not have a simple analytic form: in this case, the posterior distribution can be simulated using [[Markov chain Monte Carlo]] techniques, while optimization to find its mode(s) may be difficult or impossible.<br /> <br /> == References ==<br /> * M. DeGroot, ''Optimal Statistical Decisions'', McGraw-Hill, (1970).<br /> <br /> == External links ==<br /> * [http://en.wikipedia.org/wiki/Maximum_a_posteriori Wikipedia article on '''Maximum a posteriori''']<br /> <br /> [[Category:Academic Research]]<br /> [[Category:Statistics]]</div>

Christoph