Unsupervised Lear Ning of Gamma Mixture Models Using Minimum Message Length

  • YUDI AGUSTA Computer ScienceandSoftwareEngineering MonashUniversity Clayton, David Clayton
  • Published 2003
Mixture modellingor unsuper visedclassificationis aproblem of identifying and modelling componentsin a body of data. Earlier work in mixture modelling using Minimum MessageLength (MML) includes the multinomial and Gaussiandistributions (Wallaceand Boulton, 1968), the von MisescircularandPoissondistributions(Wallace andDowe, 1994, 2000) andthe distribution… CONTINUE READING