The 'dpmixsim' package implements a Dirichlet Process Mixture (DPM) model for clustering and image segmentation. The DPM model is a Bayesian nonparametric methodology that relies on MCMC simulations for exploring mixture models with an unknown number of components. The code implements conjugate models with normal structure (conjugate normal-normal DP mixture model). The package's applications are oriented towards the classification of magnetic resonance images according to tissue type or region of interest.
Changes in dpmixsim version 0.0-9
o Updated to remove the C keyword 'register'
Changes in dpmixsim version 0.0-8
o Correction for non-ASCII demo file
o Correction for console output
Changes in dpmixsim version 0.0-7
Corrections for partial argument match check-notes
Changes in dpmixsim version 0.0-6
o Updated data directories for compatibility with R version 2.13.0
Changes in dpmixsim version 0.0-5
o Initialisation of the simulation may be performed with any number
of clusters between 1 and n (vector data dimension).
o Simulation now estimates one variance per cluster.
o Simulation runtime have been reduced by more than 40%, based on code
optimizations.
o Argument "minvar" has been added to dpmixsim() to control the minimum admissible cluster variance estimate.
o Reversed the order of the steps in "src/gibbsclustersamplealpha.cc".
Re-sampling steps are now performed before cluster management.
o Introduced "testMarronWand" demo to test dpmixsim discriminating power.
o postdpmixciz() has been modified to go with the changes.
o Bug fixed: sampling theta in birth of "newcluster".
First version released on CRAN: 0.0-3