- 1、本文档共6页,可阅读全部内容。
- 2、原创力文档(book118)网站文档一经付费(服务费),不意味着购买了该文档的版权,仅供个人/单位学习、研究之用,不得用于商业用途,未经授权,严禁复制、发行、汇编、翻译或者网络传播等,侵权必究。
- 3、本站所有内容均由合作方或网友上传,本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺!文档内容仅供研究参考,付费前请自行鉴别。如您付费,意味着您自己接受本站规则且自行承担风险,本站不退款、不进行额外附加服务;查看《如何避免下载的几个坑》。如果您已付费下载过本站文档,您可以点击 这里二次下载。
- 4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等,请点击“版权申诉”(推荐),也可以打举报电话:400-050-0827(电话支持时间:9:00-18:30)。
查看更多
3A Vol.39 No. 3A
2011 3 ACTA ELECTRONICA SINICA Mar. 2011
BIC Gibbs
刘伟峰, 杨爱兰
( , 310018)
: , BIC Gibbs
, Gibbs , Bayes ( BIC) ,
. , , ,
.
: BIC ; Gibbs ; ;
: TP18 : A : ( 2011) 3A13406
Unsupervised Learning for Finite Mixture Models Based on
BIC Criterion and Gibbs Sampling
LIU Weife g,YANG Aila
( School of A utomation, H angz hou Dianz i University , H angz hou , Zhej i ang 310018 China )
Abstract: A u supervised lear i g algorithm for fi ite mixture models ( FMM) by usi g the BIC criterio a d the Gibbs
sampli g is proposed.The FMM parameters are estimated by the Gibbs sampli g algorithm. The umber of the models is further
give by calculati g the BIC criterio . I the fi al simulatio s, we propose two examples of Gaussia mixture models, where the
FMM with differe t umber of compo e ts is adopted to fit observatio data.The fi al result showsthat the proposed algorithm ca
effectively estimate the parameters a d the umber of the compo e ts.
Ke words: BIC criterio ; Gibbs sampli g; u supervised lear i g;fi ite mixture models
[ 5~ 9]
1 .
:
, , . ,
, 20 , , ,
1000 . ,
, , . , .
, , ,
[ 1]
. , Markov Bayes .
[ 2]
. , , , , , EM
. ( EM Bayes ) .
, . ,
5 , Markov ( RJM
., , [ 10]
CMC) , ,
Mark
文档评论(0)