Wrapper around liblinear-tools

Only L2-regularized logistic regression is supported currently. Each model is trained on balanced bootstraps from the training set (one bootstrap for the positive class, one for the negative class). The size of the bootstrap is the size of the smallest (under-represented) class. Bagging is supported and allows to obtain better models.

usage: linwrap -i <filename>: training set or DB to screen output file C w1 of bags for bagging (default=off) of cross validation random seed set portion (in [0.0:1.0]) mode; use trained models mode; save trained models for best C weight to counter class imbalance number of bags (advice: optim. k rather than w)