Install and setup

MSVMpack can be installed on either Linux, Mac OS X or Windows. Another alternative is to use MSVMpack through the platform-independent web interface (see section 2.6).

Linux and Mac OS X

To install MSVMpack, simply follow the instructions below.

1. Download and extract the archive:
   unzip MSVMpack1.5.zip
   (this should create a directory named MSVMpack1.5)
2. Go to this directory and build the programs:
   cd MSVMpack1.5
   make
   (section 3.1 gives a description of additional options for this step)
3. For a system-wide install of the software type:
   make install
   (this step requires root privileges)

This should build the MSVMpack library and generate three command-line tools:

• trainmsvm: used to train an M-SVM,
• predmsvm: used to make predictions with a trained M-SVM,
• msvmserver: the web server.

See the README file for a full description of the contents of the MSVMpack1.5 directory.

Note for Mac OS X users.

The developer tools with the optional command-line tools must be installed on the system in order to build MSVMpack. These tools can be found on the installation disk of Mac OS X or on Apple website.

Windows

Precompiled executables are provided in the MSVMpack1.5 $\backslash$Windows $\backslash$bin directory. To use these and run MSVMpack, follow the instructions below.

1. Download and extract the archive in a directory of your choice, say DIR
2. Open an MS-DOS window: Start $\rightarrow$ Execute... $\rightarrow$ cmd
3. Go to the right directory:
   cd DIR
4. Run the setup program:
   MSVMpack1.5 $\backslash$Windows $\backslash$setPath.bat
5. For a 32 bits environment, additionnally run :
   MSVMpack1.5 $\backslash$Windows $\backslash$Utils $\backslash$32bits $\backslash$setup32bits.bat
6. Close the MS-DOS window and open a new one for the setup to take effect.

Then, you can use trainmsvm.exe and predmsvm.exe as explained in the following sections. In order to benefit from special optimization options or include your own custom kernel functions, you need to recompile MSVMpack. How to build MSVMpack on Windows is detailed in Section 3.2.