Export data for use with SemiL. I would have liked to have computed the
distance using the COLT library; however this was far too slow.
Produce following files:
- [scope]_data.txt - sparse data file. Can be converted into distance
matrix for SemiL using R or Matlab. R script provided. If you want to use
semiL to generate a distance matrix, use the libsvm formatter.
- [fold]_label.txt - semiL label file, one for each fold. class labels
corresponding to rows. test data automatically unlabeled. The same
data/distance matrix can be used with different label files - the rows across
folds refer to the same instance ids. What differs is the labels for the test
instances (0 for test instance's fold).
- [fold]_class.txt - contains instance ids and target class ids for each
fold. matrix with 3 columns: instance id, train/test, target class id. Used
by SemiLEvaluationParser to evaluate SemiL predictions.