Skip to content

hpicgs/bhtsne

 
 

Repository files navigation

Build Status Build Status

This software package contains a Barnes-Hut implementation of the t-SNE algorithm as a C++ library and CLI. The implementation is described in this paper.

Requirements

  • C++14

Optional:

  • openMP (supported on all major plattforms including Solaris, AIX, HP-UX, Linux, macOS, and Windows)
  • AVX2 (first supported by Intel with the Knights Landing architecture)

Installation

This project uses the cmake-init template.

For MacOS and Linux:

./configure
cmake --build build

For Windows we recommend the cmake GUI to create a Visual Studio solution.

C++ Library Usage Example

#include <bhtsne/TSNE.h>

int main(int argc, char * argv[])
{
  bhtsne::TSNE alg;
  alg.loadCSV("..\data\small.csv");
  alg.setPerplexity(0.2);
  alg.run();
  tsne.saveSVG();
  tsne.saveCSV();
  return 0;
}

For a more information please refer to our doxygen documentation. The documentation can be enabled with the OPTION_BUILD_DOCS cmake option. Please note that linking against the library may require additional linker options for OpenMP.

CLI Usage Examples

./bhtsne_cmd --help
./bhtsne_cmd --version
./bhtsne_cmd --iterations 1000 --random-seed 42 -svg ~/data.dat
cat data.csv | ./bhtsne_cmd -legacy
./bhtsne_cmd -stdout -csv -legacy -svg ~/small.csv

Releases

No releases published

Packages

No packages published

Languages

  • C++ 84.1%
  • Python 11.0%
  • CMake 2.2%
  • Shell 0.9%
  • Makefile 0.7%
  • M4 0.6%
  • Other 0.5%