Please see the file CHANGES.md for the changelog of RGF.

We thank Dave Slate for suggesting a solution to the issue of file size restriction.

1. Introduction

   1.1. System Requirements

2. Download and Installation

3. Creating the Executable

   3.1. Windows

   3.2. Unix-like Systems

   3.3. [Optional] Endianness Consideration

4. Documentation

5. Contact

6. Copyright

7. References

This software package provides implementation of regularized greedy forest (RGF) described in [1].

Executables are provided only for some versions of Windows (see Section 3) for detail). If provided executables do not work for your environment, you need to compile C++ code.

To use the provided tools and to go through the examples in the user guide, Perl is required.

Download the package and extract the content. Otherwise, you can use git

git clone https://github.com/RGF-team/rgf.git

The top directory of the extracted (or cloned) content is rgf and this software package is located into RGF subfolder. Below all the path expressions are relative to rgf/RGF.

To go through the examples in the user guide, your executable needs to be at the bin directory (it will appear there after you compile it by any method from the listed below). Otherwise, your executable can be anywhere you like.

The easiest way. Just download the precompiled file rgf.exe from the latest GitHub release.

1. Open directory Windows/rgf.

2. Open rgf.sln file with Visual Studio.

   If you are asked to upgrade the solution file, click OK.

3. Choose Release from the Solution Configurations listbox and Win32 or x64 depending on your machine architecture from the Solution Platforms listbox on the standard toolbar.

4. Choose BUILD -> Build Solution (Ctrl+Shift+B).

   If you have errors about Platform Toolset, go to PROJECT -> Properties -> Configuration Properties -> General and select the toolset installed on your machine.

Build executable file with MinGW g++ from existing makefile (you may want to customize this file for your environment).

mkdir bin
cd build
mingw32-make

Create solution file with CMake and then compile with Visual Studio.

cd build
cmake ../ -G "Visual Studio 10 2010"
cmake --build . --config Release

If you are compiling on 64-bit machine, then add -A x64 to the end of command: cmake ../ -G "Visual Studio 10 2010" -A x64.

We tested following versions of Visual Studio:

• Visual Studio 10 2010 [Win64]
• Visual Studio 11 2012 [Win64]
• Visual Studio 12 2013 [Win64]
• Visual Studio 14 2015 [Win64]
• Visual Studio 15 2017 [Win64]
• Visual Studio 16 2019 [Win64]
• Visual Studio 17 2022 [Win64]

Other versions may work but are untested.

Create makefile with CMake and then compile with MinGW.

cd build
cmake ../ -G "MinGW Makefiles"
cmake --build . --config Release

Build executable file with g++ from existing makefile (you may want to customize this file for your environment).

mkdir bin
cd build
make

Create makefile with CMake and then compile.

cd build
cmake ../
cmake --build . --config Release

The models obtained by RGF training can be saved to files. The model files are essentially snapshots of memory that include numerical values. Therefore, the model files are sensitive to "endianness" of the environments. For this reason, if you wish to share model files among environments of different endianness, you need to follow the instructions below. Otherwise, you can skip this section.

To share model files among the environments of different endianness, build your executable for the environment with big-endian with the compile option:

/D_AZ_BIG_ENDIAN_

By doing so, the executable in your big-endian environment swaps the byte order of numerical values before writing and after reading the model files.

Please see the file rgf-guide.rst "Regularized Greedy Forest: User Guide".

Please post an issue at GitHub repository for any errors you encounter.

RGF is distributed under the MIT license. Please read the file COPYING.

[1] Rie Johnson and Tong Zhang. Learning Nonlinear Functions Using Regularized Greedy Forest. IEEE Transactions on Pattern Analysis and Machine Intelligence, 36(5):942-954, May 2014.