The current programs use:

LibSVM version 3.1 (April, 2011), available from https://www.csie.ntu.edu.tw/~cjlin/libsvm/oldfiles/libsvm-3.1.zip
Drtoolbox version: 0.8.1b (March, 2013), available from http://lvdmaaten.github.io/drtoolbox.

Both libraries are not included by copyright reasons, so the code in its actual state will not run. In order to run it:

1) Download libsvm 3.1 from https://www.csie.ntu.edu.tw/~cjlin/libsvm/oldfiles/libsvm-3.1.zip into a directory called libsvm-3.1 in directory programas. Our code might also run with newer LibVSM versions, but some minor updatings might be required. Please read the README file in the matlab directory of the libsvm-3.1 directory in order to install the Matlab interface (you just have to run 'make' from the Linux command terminal; please change properly the path where Matlab is installed on your computer in Makefile on the libsvm-3.1/matlab directory).

2) Download Drtoolbox from http://lvdmaaten.github.io/drtoolbox into a directory called drtoolbox in directory programas. 

After downloading both libraries you will be able to run the current programs.

In order to run a projection (or mapping), different to svmNd, please run Matlab, change to directory programas and execute from the command window:

proxecta(mapping,dataset)

where:
mapping is the mapping name (e.g., PKDA2, LDA, etc.), listed in file lista_proxeccions.txt, in directory programas
dataset is the dataset name (e.g., acute-inflammation, acute-nephritis, etc.), listed in file lista_problemas.txt, in directory programas

After running the mapping, the directory resultados/dataset/mapping will content five files (dataset and mapping must be replaced by the dataset and mapping names in each file name):

1) kappa.dat: the Cohen kappa (in %) achieved by LibSVM on the 2D mapped dataset.

2) mapa_libsvm_dataset_mapping.png: an image in PNG (portable network graphics) with the 2D classification map (including the 2D mapped training, validation and test patterns, and the colors showing the class to which each point in the 2D space is assigned by LibSVM). The class colors for classes are blue, red, yellow, green, white, magenta, cyan and black for classes 1, 2, 3, 4, 5, 6, 7, 8. The maximum number of classes allowed is 8. This limit is caused by the limited number of colors (also, too many colors can be difficult to distinguish).

3) dataset_2D_mapping.dat: 2D patterns for the training patterns (which are 50% of the whole patterns) in two columns. The class for each pattern is in the third column.

4) proxeccion_dataset_mapping.png: a PNG image with all the 2D mapped patterns (training, validation and test patterns).

5) resultados_libsvm_pima_PKDA2.dat: log file with all the information of the mapping and the classification of the 2D mapped patterns by LibSVM, namely the hyper-parameter tuning, the best confusion matrix during tuning and the test confusion matrix and kappa (line kappa_mellor).

------------------------------------------------------------------------------------

In order to run svmNd (the SVM classifier on the high-dimensional original data), please run from the Matlab command window:

svmNd(dataset)

where:

dataset is the dataset name (e.g., acute-inflammation, acute-nephritis, etc.), listed in file lista_problemas.txt, in directory programas

The results are stored in directory resultados/dataset/mapping, but in this case only files kappa.dat and resultados_libsvm_dataset_svmNd.dat are created, because no 2D mapping is ran.