org.apache.spark.mllib.classification
LogisticRegressionWithLBFGS
Related Doc:
package classification
org.apache.spark.mllib.classification
Whether to add intercept (default: false).
Whether to add intercept (default: false).
Create a model given the weights and intercept
Create a model given the weights and intercept
Generate the initial weights when the user does not supply them
Generate the initial weights when the user does not supply them
The dimension of training features.
The dimension of training features.
Get if the algorithm uses addIntercept
Get if the algorithm uses addIntercept
The dimension of training features.
The dimension of training features.
In GeneralizedLinearModel, only single linear predictor is allowed for both weights
and intercept.
In GeneralizedLinearModel, only single linear predictor is allowed for both weights
and intercept. However, for multinomial logistic regression, with K possible outcomes,
we are training K-1 independent binary logistic regression models which requires K-1 sets
of linear predictor.
As a result, the workaround here is if more than two sets of linear predictors are needed,
we construct bigger weights vector which can hold both weights and intercepts.
If the intercepts are added, the dimension of weights will be
(numOfLinearPredictor) * (numFeatures + 1) . If the intercepts are not added,
the dimension of weights will be (numOfLinearPredictor) * numFeatures.
Thus, the intercepts will be encapsulated into weights, and we leave the value of intercept in GeneralizedLinearModel as zero.
The optimizer to solve the problem.
The optimizer to solve the problem.
Run Logistic Regression with the configured parameters on an input RDD of LabeledPoint entries starting from the initial weights provided.
Run Logistic Regression with the configured parameters on an input RDD of LabeledPoint entries starting from the initial weights provided.
If a known updater is used calls the ml implementation, to avoid applying a regularization penalty to the intercept, otherwise defaults to the mllib implementation. If more than two classes or feature scaling is disabled, always uses mllib implementation. Uses user provided weights.
In the ml LogisticRegression implementation, the number of corrections
used in the LBFGS update can not be configured. So optimizer.setNumCorrections()
will have no effect if we fall into that route.
Run Logistic Regression with the configured parameters on an input RDD of LabeledPoint entries.
Run Logistic Regression with the configured parameters on an input RDD of LabeledPoint entries.
If a known updater is used calls the ml implementation, to avoid applying a regularization penalty to the intercept, otherwise defaults to the mllib implementation. If more than two classes or feature scaling is disabled, always uses mllib implementation. If using ml implementation, uses ml code to generate initial weights.
Set if the algorithm should add an intercept.
Set if the algorithm should add an intercept. Default false. We set the default to false because adding the intercept will cause memory allocation.
Set the number of possible outcomes for k classes classification problem in Multinomial Logistic Regression.
Set the number of possible outcomes for k classes classification problem in Multinomial Logistic Regression. By default, it is binary logistic regression so k will be set to 2.
Set if the algorithm should validate data before training.
Set if the algorithm should validate data before training. Default true.
Train a classification model for Multinomial/Binary Logistic Regression using Limited-memory BFGS. Standard feature scaling and L2 regularization are used by default. NOTE: Labels used in Logistic Regression should be {0, 1, ..., k - 1} for k classes multi-label classification problem.
Earlier implementations of LogisticRegressionWithLBFGS applies a regularization penalty to all elements including the intercept. If this is called with one of standard updaters (L1Updater, or SquaredL2Updater) this is translated into a call to ml.LogisticRegression, otherwise this will use the existing mllib GeneralizedLinearAlgorithm trainer, resulting in a regularization penalty to the intercept.