Given a feature matrix and a list of sources and targets that determines whether an instance is positive or negative, train a model.
Give a score to every row in the feature matrix, according to the learned weights.
Give a score to every row in the feature matrix, according to the learned weights.
A feature matrix specified as a list of MatrixRow
objects.
Each row receives a score from the classifier.
A map from source node to (target node, score) pairs, where the score is computed from the features in the feature matrix and the learned weights.
Handles learning and classification for models that do batch training.
Note that this only deals with _feature indices_, and has no concept of path types or anything else. So you need to be sure that the feature indices don't change between training and classification time, or your model will be all messed up.