orange.LinearLearner
is a learner that uses the LIBLINEAR library backend that is very fast on large datasets.
Linear learner learnes the attribute weights using one of the four possible methods.
Attributes
orange.LinearLearner.L2_LR (L2-regularized logistic regression, default)
orange.LinearLearner.L2LOSS_SVM_DUAL
orange.LinearLearner.L2LOSS_SVM
orange.LinearLearner.L1LOSS_SVM_DUAL
L2_LR
supports probabilty esstimations.Linear classifiers that uses one class vs. rest strategy for multi-class classification. It supports probability esstimation only if it was build with L2-regularized logistic regression learner.
Attributes