A C++ generic programming library for machine learning, developped by Frédéric Pennerath, Jérémy Fix and Hervé Frezza-Buet.
The directory contains several source packages, the documentation for each of them can be found in the doc directory as a doxygen generated html file.
For example, the documentation of gaml-libsvm is accessible from doc/gaml-libsvm/index.html
First, get the files.
git clone https://github.com/HerveFrezza-Buet/gaml
Then, you can install all packages as follows. The commands below concern the installation of the gaml package, the installation of the other packages is similar. For a 32bit architecture:
mkdir -p gaml/build
cd gaml/build
cmake .. -DCMAKE_INSTALL_PREFIX=/usr
sudo make install
cd ../..
For a Fedora-64bit architecture:
mkdir -p gaml/build
cd gaml/build
cmake .. -DCMAKE_INSTALL_PREFIX=/usr -DLIB_SUFFIX=64
sudo make install
cd ../..
The default installation uses GNU compilers gcc/g++. However clang is also supported. Just set CC and CXX environment variables as follows before running cmake.
export CC= /usr/bin/clang
export CXX= /usr/bin/clang++
The available packages are :
- gaml
- The core library. It provide generic tools for usual data handling in machine learning.
- gaml-linear
- Linear learning (LASSO and LARS) with gaml.
- gaml-libsvm
- Libsvm support. libsvm should be installed first.
- gaml-xtree
- Extreme decision trees support.
- gaml-mlp
- Multi-layer perceptron support. easykf should be installed first.
- gaml-kvq
- Kernelized vector quantization support, by linking with vq2. vq2 should be installed first. Notes that the basics for kernelized vector quantization are available in the core gaml package (the span namespace).