gregory-bopp / gmm

Gaussian Mixture Model

Gaussian Mixture Model

One goal of this project was to become more familiar with R internals. This package contains two simple implementations of an EM algorithm to fit a Gaussian mixture model, one in R and one with (stl only) c++. Finally, the c++ is ported into R without the use of Rcpp.

See mclust for a more robust R implementation for fitting Gaussian mixture models.

R Installation

devtools::install_github("gregory-bopp/gaussian_mixture_model/gmm")

1. R Implementation

See example/r_example.R for an example of simulating data from a GMM and fit using the gmm EM implementation.

2. Standalone C++ Implementation

1. Simulate GMM data with example/example.R. Data and initial values are stored in example/data/
2. Build gmm cpp program with make from gmmcpp
3. Run cpp implementation of EM algorithm to find ML estimates for GMM with

$ ./gmmcpp/bin/gmm -v -i ../example/data/init.csv -o ../example/data/fit.csv ../example/data/data.csv 

R wrapper of C++ Implementation

See example/cpp_example.R