Package jsat.classifiers.linear

Class Summary

Class	Description
ALMA2	Provides a linear implementation of the ALMAp algorithm for p = 2, which is considerably more efficient to compute.
AROW	An implementation of Adaptive Regularization of Weight Vectors (AROW), which uses second order information to learn a large margin binary classifier.
BBR	This is an implementation of Bayesian Binary Regression for L1 and L2 regularized logistic regression.
LinearBatch	LinearBatch learns either a classification or regression problem depending on the loss function ℓ(w,x) used.
LinearL1SCD	Implements an iterative and single threaded form of fast Stochastic Coordinate Decent for optimizing L1 regularized linear regression problems.
LinearSGD	LinearSGD learns either a classification or regression problem depending on the loss function ℓ(w,x) used.
LinearTools	This class provides static helper methods that may be useful for various linear models.
LogisticRegressionDCD	This provides an implementation of regularized logistic regression using Dual Coordinate Descent.
NewGLMNET	NewGLMNET is a batch method for solving Elastic Net regularized Logistic Regression problems of the form 0.5 * (1-α) ||w||2 + α * ||w||1 + C * ∑Ni=1 ℓ (wT xi + b, yi).
NHERD	Implementation of the Normal Herd (NHERD) algorithm for learning a linear binary classifier.
PassiveAggressive	An implementations of the 3 versions of the Passive Aggressive algorithm for binary classification and regression.
ROMMA	Provides an implementation of the linear Relaxed online Maximum Margin algorithm, which finds a similar solution to SVMs.
SCD	Implementation of Stochastic Coordinate Descent for L1 regularized classification and regression.
SCW	Provides an Implementation of Confidence-Weighted (CW) learning and Soft Confidence-Weighted (SCW), both of which are binary linear classifiers inspired by PassiveAggressive.
SMIDAS	Implements the iterative and single threaded stochastic solver for L1 regularized linear regression problems SMIDAS (Stochastic Mirror Descent Algorithm mAde Sparse).
SPA	Support class Passive Aggressive (SPA) is a multi class generalization of PassiveAggressive.
STGD	This provides an implementation of Sparse Truncated Gradient Descent for L1 regularized linear classification and regression on sparse data sets.
StochasticMultinomialLogisticRegression	This is a Stochastic implementation of Multinomial Logistic Regression.
StochasticSTLinearL1	This base class provides shared functionality and variables used by two different training algorithms for L1 regularized linear models.

Enum Summary

Enum	Description
BBR.Prior	Valid priors that control what type of regularization is applied
NHERD.CovMode	Sets what form of covariance matrix to use
PassiveAggressive.Mode	Controls which version of the Passive Aggressive update is used
SCW.Mode	Which version of the algorithms shuld be used
StochasticMultinomialLogisticRegression.Prior	Represents a prior of the coefficients that can be applied to perform regularization.
StochasticSTLinearL1.Loss