MeanShift_cpp is an implementation of MeanShift clustering in C++

The current version is functional, but very limited

The current programming model includes a class MeanShift. The MeanShift constructor takes a function pointer to a kernel function to be used in the clustering process. If NULL, it will use a Gaussian kernel.

Points are assumed to be vector<double>. Point collections are assumed to be vector<vector<double> >. This may change in the future.

To cluster a set of points, create a MeanShift object and call the cluster method with a collection of points and a kernel bandwidth. The shifted point will be returned as a vector<vector<double> >. Note that the points are not currently grouped into cluters, but rather shifted to their converged locations.

vector<vector<double> > points = load_points("test_simple.csv");
MeanShift *ms = new MeanShift(NULL);
double kernel_bandwidth = 2;
vector<vector<double> > shifted_points = ms->cluster(points, 2);

Install gnuplot and gnuplot-qt

sudo apt-get install gnuplot gnuplot-qt

Make and run the cpp_test

make
./MeanShift

The program will generate a csv file named "result.csv".

Plot it

gnuplot
plot 'test.csv' with points, 'result.csv' with points