Parallel Iterative Re-Weighted Least Squares: A Parallel Support Vector Machine (SVM) solver based on the IRWLS algorithm

This software is implemented in C and requires the following libraries:

• [OpenMP] (http://openmp.org/wp/) To parallelize the software
• A Linear Algebra Package that implements the BLAS and Lapack standard routines, this software has been tested with these libraries (you need just one of them):
  • [BLAS] (http://www.netlib.org/blas/) and [LAPACK] (http://www.netlib.org/lapack/)
  • [ATLAS] (http://math-atlas.sourceforge.net/)
  • MKL

External libraries:

An example to install in ubuntu all the libraries that we need is the following:

OPENMP is currently included with the gcc compiler, if gcc is not installed, use the following command line:

sudo apt-get install build-essential

To install the linear algebra routines use the following command line:

sudo apt-get install liblapack-dev
sudo apt-get install libblas-dev
sudo apt-get install libatlas-dev

Compiling:

If you use blas, lapack, atlas:

make

If you use MKL libraries:

make USE_MKL=1

If the libraries are not installed in the standard paths you can uncomment and edit the variables INCLUDEPATH and LIBRARYPATH to tell the location.

Training:

To train the algorithm and create the model:

./PIRWLS-train [options] training_set_file model_file

Options:

• -g Gamma: Set gamma in the radial basis kernel function (default 1)
• -c Cost: Set the SVM Cost (default 1)
• -w Working_set_size: Size of the Least Squares Problem in every iteration (default 500)
• -t Number_of_Threads: It is the number of threads in the parallel task (default 1)

Example:

./PIRWLS-train -g 0.001 -c 1000 -t 4 training_set_file.txt model_file.mod

Test:

To make predictions with the model in a different dataset:

./PIRWLS-predict [options] dataset_file model_file output_file

Options:

• -t Number_of_Threads: It is the number of threads in the parallel task (default 1)
• -l Labeled: (default 0)
  • 1 if the dataset is labeled (shows accuracy)
  • 0 if the dataset is unlabeled

Example:

./PIRWLS-predict -t 4 -l 1 dataset_file.txt model_file.mod output_file.txt

The dataset must be provided in LibSVM format, labeled to train the model and labeled or unlabeled for predictions (using the -l option in the PIRWLS-predict command to tell if the file is labeled or unlabeled):

Labeled example:

+1 1:5 7:2 15:6
+1 1:5 7:2 15:6 23:1
-1 2:4 3:2 10:6 11:4

Unlabeled example:

1:5 7:2 15:6
1:5 7:2 15:6 23:1
2:4 3:2 10:6 11:4