Package LBJ2.learn

Learning algorithms, normalizers (used in inference; see Normalizer), testing metrics (used in cross validation; see TestingMetric), and other utility classes can be found in this package.

See:
Description

Interface Summary

TestingMetric

TestingMetric is an interface through which the user may implement their own testing method for use by LBJ's internal cross validation algorithm.

Class Summary

Accuracy	This is the cross validation testing metric which LBJ defaults to when none is specified.
AdaBoost	Implementation of the AdaBoost binary classification learning algorithm.
BinaryMIRA	The Binary MIRA learning algorithm implementation.
BinaryMIRA.Parameters	Simply a container for all of BinaryMIRA's configurable parameters.
IdentityNormalizer	This Normalizer simply returns the same ScoreSet it was passed as input without modifying anything.
Learner	Extend this class to create a new Classifier that learns to mimic one an oracle classifier given a feature extracting classifier and example objects.
LearnerToText	This extremely simple class can be used to print a textual representation of a trained learner to STDOUT.
LinearThresholdUnit	A LinearThresholdUnit is a Learner for binary classification in which a score is computed as a linear function a weight vector and the input example, and the decision is made by comparing the score to some threshold quantity.
LinearThresholdUnit.Parameters	Simply a container for all of LinearThresholdUnit's configurable parameters.
Log	Simply turns each score s in the ScoreSet returned by the specified Normalizer into log(s).
MultiLabelLearner	A simple implementation of a learner that learns from examples with multiple labels and is capable of predicting multiple labels on new examples.
MultiLabelLearner.Parameters	Simply a container for all of MultiLabelLearner's configurable parameters.
MuxLearner	A MuxLearner uses one of many Learners indexed by the first feature in an example to produce a classification.
MuxLearner.Parameters	Simply a container for all of MuxLearner's configurable parameters.
NaiveBayes	Naive Bayes is a multi-class learner that uses prediction value counts and feature counts given a particular prediction value to select the most likely prediction value.
NaiveBayes.Count	A Count object stores two doubles, one which holds a accumulated count value and the other intended to hold the natural logarithm of the count.
NaiveBayes.Parameters	Simply a container for all of NaiveBayes's configurable parameters.
Normalizer	A normalizer is a function of a ScoreSet producing normalized scores.
Parameters	Parameters classes are used to hold values for learning algorithm parameters, and all learning algorithm implementations must provide a constructor that takes such an object as input.
Sigmoid	The sigmoid normalization function replaces each score xi with 1 / (1 + exp(-alpha xi)), where alpha is a user-specified constant.
Softmax	The softmax normalization function replaces each score with the fraction of its exponential out of the sum of all scores' exponentials.
SparseAveragedPerceptron	An approximation to voted Perceptron, in which a weighted average of the weight vectors arrived at during training becomes the weight vector used to make predictions after training.
SparseAveragedPerceptron.AveragedWeightVector	This implementation of a sparse weight vector associates two doubles with each Feature.
SparseAveragedPerceptron.Parameters	Simply a container for all of SparseAveragedPerceptron's configurable parameters.
SparseNetworkLearner	A SparseNetworkLearner uses multiple LinearThresholdUnits to make a multi-class classification.
SparseNetworkLearner.Parameters	Simply a container for all of SparseNetworkLearner's configurable parameters.
SparsePerceptron	Simple sparse Perceptron implementation.
SparsePerceptron.Parameters	Simply a container for all of SparsePerceptron's configurable parameters.
SparseWeightVector	This class is used as a weight vector in sparse learning algorithms.
SparseWinnow	Simple sparse Winnow implementation.
SparseWinnow.Parameters	Simply a container for all of SparseWinnow's configurable parameters.
StochasticGradientDescent	Gradient descent is a batch learning algorithm for function approximation in which the learner tries to follow the gradient of the error function to the solution of minimal error.
StochasticGradientDescent.Parameters	Simply a container for all of StochasticGradientDescent's configurable parameters.
WekaWrapper	Translates LBJ's internal problem representation into that which can be handled by WEKA learning algorithms.
WekaWrapper.Parameters	Simply a container for all of WekaWrapper's configurable parameters.

Package LBJ2.learn Description

Learning algorithms, normalizers (used in inference; see Normalizer), testing metrics (used in cross validation; see TestingMetric), and other utility classes can be found in this package.

Learning algorithms are always associated with an evaluation algorithm that actually does the classifying; the learning algorithm simply sets the parameters of that classifying function. Thus, they are implemented here as Classifiers that can change their representations (i.e., their parameters) given training data.