markov chain monte carlo methods

markov chain monte carlo methods

ID:7279952

大小:613.61 KB

页数:36页

时间:2018-02-10

markov chain monte carlo methods_第1页
markov chain monte carlo methods_第2页
markov chain monte carlo methods_第3页
markov chain monte carlo methods_第4页
markov chain monte carlo methods_第5页
资源描述:

《markov chain monte carlo methods》由会员上传分享,免费在线阅读,更多相关内容在工程资料-天天文库

1、6MarkovChainMonteCarloMethods6.1IntroductionInChapter5,weintroducedtheuseofsimulationinBayesianinference.Rejec-tionsamplingisageneralmethodforsimulatingfromanarbitraryposteriordistribution,butitcanbedifficulttosetupsinceitrequirestheconstruc-tionofasuita

2、bleproposaldensity.ImportancesamplingandSIRalgorithmsarealsogeneral-purposealgorithms,buttheyalsorequireproposaldensitiesthatmaybedifficulttofindforhigh-dimensionalproblems.Inthischapter,weillustratetheuseofMarkovchainMonteCarlo(MCMC)algorithmsinsummarizi

3、ngposteriordistributions.Markovchainsareintroducedinthedis-cretestatespacesituationinSection6.2.Throughasimplerandomwalkexample,weillustratesomeoftheimportantpropertiesofaspecialMarkovchain,andweuseRtosimulatefromthechainandmovetowardthesta-tionarydist

4、ribution.InSection6.3,wedescribetwovariantsofthepopularMetropolis-HastingsalgorithmsinsettingupMarkovchains,andinSection6.4wedescribeGibbssampling,wheretheMarkovchainissetupthroughtheconditionaldistributionsoftheposterior.Wedescribeonestrategyforsum-ma

5、rizingaposteriordistributionandillustrateitforthreeproblems.MCMCalgorithmsareveryattractiveinthattheyareeasytosetupandprogramandrequirerelativelylittlepriorinputfromtheuser.Risaconvenientlanguageforprogrammingthesealgorithmsandisalsoverysuitableforperf

6、ormingout-putanalysis,whereonedoesseveralgraphicalandnumericalcomputationstocheckifthealgorithmisindeedproducingdrawsfromthetargetposteriordistribution.6.2IntroductiontoDiscreteMarkovChainsSupposeapersontakesarandomwalkonanumberlineonthevalues1,2,3,4,5

7、,6.Ifthepersoniscurrentlyataninteriorvalue(2,3,4,or5),inthenextsecondsheisequallylikelytoremainatthatnumberormovetoanadjacentJ.Albert,BayesianComputationwithR,UseR,DOI10.1007/978-0-387-92298-06,©SpringerScience+BusinessMedia,LLC20091186MarkovChainMonte

8、CarloMethodsnumber.Ifshedoesmove,sheisequallylikelytomoveleftorright.Ifthepersoniscurrentlyatoneoftheendvalues(1or6),inthenextsecondsheisequallylikelytostaystillormovetotheadjacentlocation.ThisisasimpleexampleofadiscreteMarkovchain.AMar

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。