Expectation Maximization of Gaussian Mixture Models via CUDA: gmm_example.m - File Exchange

Code covered by the BSD License

Highlights from
Expectation Maximization of Gaussian Mixture Models via CUDA

errorellipse(mu,sigma,stdev,n...
gmm_test(num_clusts, num_dims... GMM_TEST - runs the EM algorithm on random data on the GPU
nvmex(varargin) MEX Compile MEX-function
nvmex_helper(varargin) MEX_HELPER is a helper function that contains the code that MEX.M (an
compile.m
experiment1.m
experiment2.m
experiment3.m
gmm_example.m
View all files

from Expectation Maximization of Gaussian Mixture Models via CUDA by Andrew Harp
CUDA enabled parallel EM for Gaussian Mixture Models, providing over 100x performance increases.

gmm_example.m

%GMM_EXAMPLE - simple test to show GMM EM running on GPU, plotting 
% distributions afterwards.
%
% Andrew Harp (andrew.harp@gmail.com)
% http://andrewharp.com/gmmcuda

num_clusts = 3
num_dims = 2
num_data = 3000
num_steps = 10
do_cpu = 0       % Don't bother running CPU version, just do GPU.
show_plot = 1

% Run the example clustering!
gmm_test(num_clusts, num_dims, num_data, num_steps, do_cpu, show_plot);

Highlights from Expectation Maximization of Gaussian Mixture Models via CUDA

Highlights from
Expectation Maximization of Gaussian Mixture Models via CUDA