org.apache.flink.ml.classification.logisticregression

类 LogisticGradient

java.lang.Object

org.apache.flink.ml.classification.logisticregression.LogisticGradient

所有已实现的接口:

Serializable

public class LogisticGradient
extends Object
implements Serializable

Utility class to compute gradient and loss for logistic loss function.

See http://mlwiki.org/index.php/Logistic_Regression.

另请参阅:

序列化表格

构造器概要

构造器

构造器和说明
LogisticGradient(double l2)

方法概要

限定符和类型	方法和说明
void	computeGradient(List<org.apache.flink.ml.common.feature.LabeledPointWithWeight> dataPoints, org.apache.flink.ml.linalg.DenseVector coefficient, org.apache.flink.ml.linalg.DenseVector cumGradient) Computes gradient on a set of samples.
org.apache.flink.api.java.tuple.Tuple2<Double,Double>	computeLoss(List<org.apache.flink.ml.common.feature.LabeledPointWithWeight> dataPoints, org.apache.flink.ml.linalg.DenseVector coefficient) Computes weight sum and loss sum on a set of samples.

从类继承的方法 java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

构造器详细资料

LogisticGradient

public LogisticGradient(double l2)

方法详细资料

computeLoss

public org.apache.flink.api.java.tuple.Tuple2<Double,Double> computeLoss(List<org.apache.flink.ml.common.feature.LabeledPointWithWeight> dataPoints,
                                                                         org.apache.flink.ml.linalg.DenseVector coefficient)

Computes weight sum and loss sum on a set of samples.

参数:

dataPoints - A sample set of train data.

coefficient - The model parameters.

返回:

Weight sum and loss sum of the input data.

computeGradient

public void computeGradient(List<org.apache.flink.ml.common.feature.LabeledPointWithWeight> dataPoints,
                            org.apache.flink.ml.linalg.DenseVector coefficient,
                            org.apache.flink.ml.linalg.DenseVector cumGradient)

Computes gradient on a set of samples.

参数:

dataPoints - A sample set of train data.

coefficient - The model parameters.

cumGradient - The accumulated gradients.