返回
(Monte Carlo Hidden Markov Models Learning Non-Parametric Models of Partially Observable Stochastic Processes)英文文献资料

(Monte Carlo Hidden Markov Models Learning Non-Parametric Models of Partially Observable Stochastic Processes)英文文献资料

ID：39205409

大小：1.87 MB

页数：11页

时间：2019-06-27

(Monte Carlo Hidden Markov Models Learning Non-Parametric Models of Partially Observable Stochastic Processes)英文文献资料_第1页
(Monte Carlo Hidden Markov Models Learning Non-Parametric Models of Partially Observable Stochastic Processes)英文文献资料_第2页
(Monte Carlo Hidden Markov Models Learning Non-Parametric Models of Partially Observable Stochastic Processes)英文文献资料_第3页
(Monte Carlo Hidden Markov Models Learning Non-Parametric Models of Partially Observable Stochastic Processes)英文文献资料_第4页
(Monte Carlo Hidden Markov Models Learning Non-Parametric Models of Partially Observable Stochastic Processes)英文文献资料_第5页
资源描述：

《(Monte Carlo Hidden Markov Models Learning Non-Parametric Models of Partially Observable Stochastic Processes)英文文献资料》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、MonteCarloHiddenMarkovModels:LearningNon-ParametricModelsofPartiallyObservableStochasticProcessesSebastianThrunJohnC.LangfordDieterFoxRobotLearningLaboratory,SchoolofComputerScienceCarnegieMellonUniversity5000ForbesAve.,Pittsburgh,PA15213fthrun,jcl,dfoxg@cs.cmu.eduAbstractlearnnon-parametr

2、ichiddenMarkovmodelswithcontin-uousstateandobservationspaces.Sincecontinuousstatespacesaresufficientlyrichtooverfitanydataset,ourap-Wepresentalearningalgorithmfornon-parametricproachusesshrinkageasamechanismforregularization.hiddenMarkovmodelswithcontinuousstateandTheshrinkagefactor,whichdet

3、erminestheeffectivecapac-observationspaces.Allnecessaryprobabilityden-ityoftheHMM,isannealeddownovermultipleiterationssitiesareapproximatedusingsamples,alongwithofEM,andearlystoppingisappliedformodelselection.densitytreesgeneratedfromsuchsamples.AMonteCarloversionofBaum-Welch(EM)isem-MCHMM

4、spossessthefollowingfourcharacteristics,ployedtolearnmodelsfromdata.RegularizationwhichdistinguishesthemfromthetraditionalHMMap-duringlearningisachievedusinganexponentialproach[27]:shrinkingtechnique.Theshrinkagefactor,whichdeterminestheeffectivecapacityofthelearningal-1.Real-valuedspaces.

5、Boththestatespaceandtheob-gorithm,isannealeddownovermultipleiterationsservationspaceofMCHMMsarecontinuous.ThisisofBaum-Welch,andearlystoppingisappliedtoimportantindomainswherethetruestateandobser-selecttherightmodel.Oncetrained,MonteCarlovationspaceoftheenvironmentiscontinuous.TheHMMscanbe

6、runinanany-timefashion.Weproveimportanceofcontinuous-valuedspaceshasbeenrec-thatundermildassumptions,MonteCarloHiddenognizedbyseveralauthors,whichhaveproposedreal-MarkovModelsconvergetoalocalmaximuminvaluedextensionsusingparametricmodelssuchaslikelihoodspace,justlikeconventionalHMMs.InGaus

7、siansandneuralnetworks[2,8,9,13,17].addition,weprovideempiricalresultsobtainedina2.Non-parametric.MostexistingHMMmodelsrelyongesturerecognitiondomain.parametricdensities,definedbyasmallsetofparam-eters(discretedistributionsincluded).Thisisclearlyappropriatewhen

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。