AntMAN: Anthology of Mixture ANalysis tools

AntMan is a R package to fit Finite Bayesian Mixture model with random number of component. The MCMC algorithm beyond batman is based on point processes and offer a more computational efficeint alternative to Reversible Jump. Different mixture kernels can be specified: Univariate Gaussian, Univariate Poisson, Univariate Binomial, Multivariate Gaussian, Multivariate Bernoulli (Latent Class Analysis). For the parameters characterising the mixture kernel, we specify conjugate priors, with possibly user specified hyper-parameters. We allow for different choices for the prior on the number of components: Shifted Poisson, Negative Binomial, and Point Masses (i.e. mixtures with fixed number of components).

How to Install

With CRAN

install.packages("AntMAN")

With Github

First make sure the package is not already installed.

remove.packages("AntMAN")

Then you can use the following:

install.packages("devtools")
devtools::install_github("bbodin/AntMAN", subdir="AntMAN")

How to test

library("AntMAN")
data(galaxy, package = "AntMAN")
y_uvn = galaxy
mixture_uvn_params = AntMAN::AM_mix_hyperparams_uninorm  (m0=20.83146, k0=0.3333333, nu0=4.222222, sig02=3.661027)

mcmc_params        = AntMAN::AM_mcmc_parameters(niter=20000, burnin=5000, thin=10, verbose=1)
components_prior   = AntMAN::AM_mix_components_prior_pois (init=3,  a=1, b=1) 
weights_prior      = AntMAN::AM_mix_weights_prior_gamma(init=2, a=1, b=1)

fit <- AntMAN::AM_mcmc_fit(
  y = y_uvn,
  mix_kernel_hyperparams = mixture_uvn_params,
  mix_components_prior =components_prior,
  mix_weight_prior = weights_prior,
  mcmc_parameters = mcmc_params)

summary (fit)
plot (fit)

Package Philosophy

The main function of the AntMAN package is AM_mcmc_fit that performs a Gibbs sampling in order to estimate a mixture of a predifined type mix_kernel_hyperparams and that models a particular population given a sample y. Additionnaly a particular component prior mix_components_prior and a weight prior mix_weight_prior can be specified and mcmc_parameters will define the MCMC parameters for the Gibbs sampler (number of interation, burn-in, ...). A prior on the number of cluster (init_K) or a specific allocation (init_clustering) can also be specify. Otherwise, the default allocation we assign a different cluster for each element of the sample y.

For example, in order to identify clusters over a population of patients given a set of medical assumption:

mcmc = AM_mcmc_parameters(niter=20000)
mix  = AM_mix_hyperparams_multiber ()
fit = AM_mcmc_fit (mix, mcmc)
summary (fit)

In this example AM_mix_hyperparams_multiber is one of the possible mixture to identify. AntMAN currently support five different mixtures :

AM_mix_hyperparams_unipois(alpha0, beta0) 
AM_mix_hyperparams_uninorm(m0, k0, nu0, sig02) 
AM_mix_hyperparams_unibin(a0, b0, mb) 
AM_mix_hyperparams_multiber(a0, b0) 
AM_mix_hyperparams_multinorm(mu0, ka0, nu0, Lam0)

Additionnaly, there is three prior_component available :

AM_mix_components_prior_pois
AM_mix_components_prior_negbin
AM_mix_components_prior_dirac

For example, in the context of image segmentation, where a maximal number of colour is require, a prior dirac can be used :

mcmc = AM_mcmc_parameters(niter=20000)
mix  = AM_mix_hyperparams_multi_norm ()
prior_component = AM_mix_components_prior_dirac(10) # nothing more than 10 colours
fit = AM_mcmc_fit (mix, prior_component, mcmc)
summary (fit)

Acknowledgement

Thanks for your contributions:

David B. Dahl (@dbdahl)
Andrea Cremaschi (@AndCre87)
Abhinav V. Natarajan (@abhi01nat)

TODOs

Check all the comments in the code
File writing feature
Arrange layout of summary

To Remember before submitting

Package testing

make check has to pass locally and remotely (on travis) with no WARNINGs and no ERRORs.
Making sure to go through the points
here: https://cran.r-project.org/web/packages/submission_checklist.html
and here: https://cran.r-project.org/web/packages/policies.html
Checking that the package works for windows with no ERRORS no WARNINGS and possibly no NOTES. URL: https://win-builder.r-project.org/

Feedback from CRAN

Please replace \% in your Rd-files by %.
Please add and explain the returned objects.
Please ensure that your functions do not modify (save or delete) the user's home filespace in your examples/vignettes/tests. That is not allow by CRAN policies. Please only write/save files if the user has specified a directory.
Please replace cat() by message() or warning() in your functions (except for print() and summary() functions). Messages and warnings can be suppressed if needed.
Please always write TRUE and FALSE instead of T and F.
Check every headers with wrong authors
Please remove the command 'rm(list=ls())' from your files.

Submit the Package

The URL is here: https://cran.r-project.org/submit.html
To make the package: make AntMAN_1.1.0.tar.gz