Software used to classify and score screens for Styles 2016.

To run the classifier, you will need:

• a working version of Matlab installed and accessible in your path.
• gunzip installed to uncompress the training_set.csv.gz file.

Unzip the training set by running gunzip data/training_set.csv.gz

To train the classifier:

• run src/RAD52_train.m in a Matlab session
• Or from the command line, run matlab -nosplash -nodesktop -r src/RAD52_train.m

To train the model and produce a ROC figure on held out training data:

• run src/trainingRoc.m in a Matlab session

N.B: If you cannot get the libsvm bindings to work, you may need to follow the instructions in libsvm-mat-3.0-1/README

styles2016
|
|- data/          # contains the training set for the foci classifier.
|
|- src/           # source code for the foci classifier.
|
|- models/        # .mat model file of the classifier used to score screens.