A subinstance of TrainingData whose labels are -1 or 1.
A CandidatePool is an abstraction for a "pool" of comparable objects.
A CandidatePoolCorpus is a set of candidate pools (see CandidatePool, above).
A CandidatePoolCorpus is a set of candidate pools (see CandidatePool, above).
the pools in the corpus
Maps feature names to integers.
Maps feature names to integers. Useful for serializing TrainingData instances for consumption by command-line machine learning tools.
an indexed sequence of feature names
The name of a feature, represented as a list of Symbols.
The name of a feature, represented as a list of Symbols.
the list of symbols comprising the feature name
A mapping from feature names to values.
A mapping from feature names to values.
Unspecified feature names are assumed to correspond to a value of zero.
the map from feature names to values
A weighted linear combination of features.
A weighted linear combination of features.
map from feature names to weight coefficients
Abstraction for a set of labeled feature vectors.
Abstraction for a set of labeled feature vectors.
Provides various serialization options for different machine learning tools.
a sequence of feature vectors labeled with doubles
A CandidatePool is an abstraction for a "pool" of comparable objects.
Each object is represented as a feature vector, and is associated with a cost. Lower-cost feature vectors are "better." Usually we want to use a corpus of these "pools" to learn to distinguish "good" feature vectors from "bad" feature vectors.
a mapping of feature vectors to costs (lower is better)