统计与进化学习：隐马尔可夫模型与结构化分类方法.pdfVIP

COMP90051StatisticalandEvolutionary

Learning

Semester2,2011

Lecturer:JeffreyChanBenRubinstein

tionary

ein

HiddenMarkovModels

els

Structuredclassification

•Todate,wehavealwaysconsideredeachinstancetly,

butinmanytasks,thereis“structure”weeninstances,e.g.:

⋆sequentialstructure(e.g.timeseriesanalysis)

⋆hierarchicalstructure(e.g.classifyingwebpageswithinaweb

site)

⋆graphstructure(e.g.derivingan“influencematrix”forasocial

network)

•Thiscallsforstructuredclassificationmodelswhichareableto

capturetheinteractionweeninstances

ndependently,

nces,e.g.:

withina

web”fora

social

hareableto