Requirements: ============= - A somewhat recent C/C++ compiler, see the list below for tested compiler/operating system combinations. - Linux: - The bash shell needs to be installed (this is often the default shell, anyway). - Windows: - A 64bit version of Windows - Optional: Visual Studio 2015, if you are going to recompile the programs for increased speed. So far the package has been successfully compiled on the following operating systems: NAME VERSION COMPILER Debian 7 GCC/G++ 4.7.2 Debian 7 GCC/G++ 4.9.0 Debian 8.7 clang++ 3.5.0 (by `make CPPC=clang++`) Ubuntu 12.05.05 GCC/G++ 4.6.2 Kubuntu 14.04 GCC/G++ 4.8.2 Fedora 22 GCC/G++ 5.1.1 OS X 10.10.3 Apple LLVM 7.0.2 Windows 8.1 Enterprise Visual Studio 2015 Windows 8.1 GCC/G++ 4.9.3 (MinGW/Rtools 3.4) Various other combinations may also be fine ... Installation: ============= Linux/OS X ---------- Please make sure that you have a recent version of gcc/g++ (Linux) or llvm/clang (OS X) is installed. Then follow these steps: 1. Download and save the file liquidSVM.tar.gz into the folder you wish to install the package. 2. Open a terminal window and switch to the folder liquidSVM.tar.gz has been saved to. 3. Unpack liquidSVM.tar.gz file by typing tar xzf liquidSVM.tar.gz 4. Compile the package by typing cd liquidSVM make all 5. If the compilation was successful then you can test the installation by typing ./run-demo.sh If the installation was successful, you should see a couple of training and testing runs as well as a very short introduction to the package. 6. Move the datasets you wish to work on into the folder liquidSVM/data. 7. [Optional] You may wish to add the two sub-folders scripts bin to your $PATH variable. Please consult the documentation of your operating system for the required steps. Windows ------- 1. Download and save the file liquidSVM.zip into the folder you wish to install the package. To obtain maximal speed, choose the instruction set to be as best as possible for your system. All CPUs produced in the last 10 years support sse2. Most modern systems should at least support AVX, and some even AVX2. If you are not sure what your CPU supports just try them out. 2. Open Explorer and extract liquidSVM.zip. 3. Open a terminal window and switch to the folder liquidSVM. Then test the installation by typing run-demo.bat If the installation was successful, you should see a couple of training and testing runs as well as a very short introduction to the package. If the demo seems to hang during the first example, then your anti-virus software may be the reason. We experienced this problem a couple of times and resolved it by deactivating the automatic scan of started programs. 4. Move the datasets you wish to work on into the folder liquidSVM/data. 5a. [Optional] If you wish to compile the executables from scratch, start Visual Studio and load the file ./liquidSVM.sln. Then, in Visual Studio choose the 'release' configuration and keep the x64 platform as the code does not compile under Windows for 32bit platforms. Then compile the three svm-xxx projects separately. If you are also interested in the few additional commandline tools, compile them, too. Note that for these no particular instruction set is pre-configured. 5b. [Optional] Compiling with MinGW/Rtools [Optional] i. Download and install Rtools from: https://cran.r-project.org/bin/windows/Rtools/Rtools34.exe We assume that it is installed in directory %RTOOLS% (e.g. C:\Rtools). ii. On the command line in the liquidSVM directory and with the path as above do path=%RTOOLS%\bin;%RTOOLS%\mingw_64\bin;%path% make NOTE: * The make will give some warnings (uname and/or chmod are not known; some uninitialized values...): it should be fine to ignore those. * In the instructions we set the path temporarily to include the Rtools. After that it is not used anymore, so you do not have to put it permanently into your PATH (it does not hurt to have it there though).