SVD decision surface for training and test data. Decision boundary for vector/SVD model on the space of positive and negative scores π and ν. Red and green dots represent negative and positive abstracts in the learning data set, respectively. Blue dots represent unlabeled test set abstracts. SVD, singular value decomposition.
Abi-Haidar et al. Genome Biology 2008 9(Suppl 2):S11 doi:10.1186/gb-2008-9-s2-s11