# Rh/Fragment 137 05

## < Rh

32.387Seiten in
diesem Wiki

 Typus Verschleierung Bearbeiter Hindemith Gesichtet
Untersuchte Arbeit:
Seite: 137, Zeilen: 5-13
Quelle: Rasmussen et al. 2007
Seite(n): 3, Zeilen: 22ff
We describe an approach to the problem of automatically determine [sic] the number patterns L based on the theory of infinite Gaussian mixtures or Dirichlet process mixtures. This theory is based on the observation that the mathematical limit of an infinite number of components in an ordinary finite mixture model (i.e. the patterns L in the d-dimensional model) corresponds to a Dirichlet process prior. In an infinite Gaussian mixture model there is no need to make arbitrary choices about how many patterns L there are in the process. The major advantage is that, although in theory the infinite mixture model has an infinite number of parameters, it is possible to do exact inference in these infinite mixture models efficiently using Markov chain Monte Carlo (MCMC) methodology. We describe an approach to the problem of automatically clustering microarray gene expression profiles based on the theory of infinite Gaussian mixtures (or Dirichlet process mixtures (DPM)) [5], [6]. This theory is based on the observation that the mathematical limit of an infinite number of components in an ordinary finite mixture model (i.e. clustering model) corresponds to a Dirichlet process prior [5]–[7]. In an infinite Gaussian mixture model there is no need to make arbitrary choices about how many clusters there are in the data. Although in theory the infinite mixture model has an infinite number of parameters, surprisingly, it is possible to do exact inference in these infinite mixture models efficiently using Markov chain Monte Carlo (MCMC) methodology, [...]
 Anmerkungen Die Quelle wird nicht erwähnt. Zur Anpassung des Untersuchungsgegenstandes werden einige Stichwortersetzungen vorgenommen. Dabei scheinen sich im ersten Satz sprachliche Fehler eingeschlichen zu haben. Sichter (Hindemith), KnallErbse