public class LinearRegression extends Regressor<Vector,LinearRegression,LinearRegressionModel> implements LinearRegressionParams, DefaultParamsWritable, org.apache.spark.internal.Logging
The learning objective is to minimize the specified loss function, with regularization. This supports two kinds of loss: - squaredError (a.k.a squared loss) - huber (a hybrid of squared error for relatively small errors and absolute error for relatively large ones, and we estimate the scale parameter from training data)
This supports multiple types of regularization: - none (a.k.a. ordinary least squares) - L2 (ridge regression) - L1 (Lasso) - L2 + L1 (elastic net)
The squared error objective function is:
$$ \begin{align} \min_{w}\frac{1}{2n}{\sum_{i=1}^n(X_{i}w - y_{i})^{2} + \lambda\left[\frac{1-\alpha}{2}{||w||_{2}}^{2} + \alpha{||w||_{1}}\right]} \end{align} $$
The huber objective function is:
$$ \begin{align} \min_{w, \sigma}\frac{1}{2n}{\sum_{i=1}^n\left(\sigma + H_m\left(\frac{X_{i}w - y_{i}}{\sigma}\right)\sigma\right) + \frac{1}{2}\lambda {||w||_2}^2} \end{align} $$
where
$$ \begin{align} H_m(z) = \begin{cases} z^2, & \text {if } |z| < \epsilon, \\ 2\epsilon|z| - \epsilon^2, & \text{otherwise} \end{cases} \end{align} $$
Since 3.1.0, it supports stacking instances into blocks and using GEMV for better performance. The block size will be 1.0 MB, if param maxBlockSizeInMB is set 0.0 by default.
Note: Fitting with huber loss only supports none and L2 regularization.
| Constructor and Description |
|---|
LinearRegression() |
LinearRegression(String uid) |
| Modifier and Type | Method and Description |
|---|---|
IntParam |
aggregationDepth()
Param for suggested depth for treeAggregate (>= 2).
|
LinearRegression |
copy(ParamMap extra)
Creates a copy of this instance with the same UID and some extra params.
|
DoubleParam |
elasticNetParam()
Param for the ElasticNet mixing parameter, in range [0, 1].
|
DoubleParam |
epsilon()
The shape parameter to control the amount of robustness.
|
BooleanParam |
fitIntercept()
Param for whether to fit an intercept term.
|
static LinearRegression |
load(String path) |
Param<String> |
loss()
The loss function to be optimized.
|
static int |
MAX_FEATURES_FOR_NORMAL_SOLVER()
When using
LinearRegression.solver == "normal", the solver must limit the number of
features to at most this number. |
DoubleParam |
maxBlockSizeInMB()
Param for Maximum memory in MB for stacking input data into blocks.
|
IntParam |
maxIter()
Param for maximum number of iterations (>= 0).
|
static MLReader<T> |
read() |
DoubleParam |
regParam()
Param for regularization parameter (>= 0).
|
LinearRegression |
setAggregationDepth(int value)
Suggested depth for treeAggregate (greater than or equal to 2).
|
LinearRegression |
setElasticNetParam(double value)
Set the ElasticNet mixing parameter.
|
LinearRegression |
setEpsilon(double value)
Sets the value of param
epsilon. |
LinearRegression |
setFitIntercept(boolean value)
Set if we should fit the intercept.
|
LinearRegression |
setLoss(String value)
Sets the value of param
loss. |
LinearRegression |
setMaxBlockSizeInMB(double value)
Sets the value of param
maxBlockSizeInMB. |
LinearRegression |
setMaxIter(int value)
Set the maximum number of iterations.
|
LinearRegression |
setRegParam(double value)
Set the regularization parameter.
|
LinearRegression |
setSolver(String value)
Set the solver algorithm used for optimization.
|
LinearRegression |
setStandardization(boolean value)
Whether to standardize the training features before fitting the model.
|
LinearRegression |
setTol(double value)
Set the convergence tolerance of iterations.
|
LinearRegression |
setWeightCol(String value)
Whether to over-/under-sample training instances according to the given weights in weightCol.
|
Param<String> |
solver()
The solver algorithm for optimization.
|
BooleanParam |
standardization()
Param for whether to standardize the training features before fitting the model.
|
DoubleParam |
tol()
Param for the convergence tolerance for iterative algorithms (>= 0).
|
String |
uid()
An immutable unique ID for the object and its derivatives.
|
Param<String> |
weightCol()
Param for weight column name.
|
featuresCol, fit, labelCol, predictionCol, setFeaturesCol, setLabelCol, setPredictionCol, transformSchemaparamsequals, getClass, hashCode, notify, notifyAll, toString, wait, wait, waitgetEpsilon, validateAndTransformSchemaextractInstances, extractInstancesgetLabelCol, labelColfeaturesCol, getFeaturesColgetPredictionCol, predictionColclear, copyValues, defaultCopy, defaultParamMap, explainParam, explainParams, extractParamMap, extractParamMap, get, getDefault, getOrDefault, getParam, hasDefault, hasParam, isDefined, isSet, paramMap, params, set, set, set, setDefault, setDefault, shouldOwntoStringgetRegParamgetElasticNetParamgetMaxItergetFitInterceptgetStandardizationgetWeightColgetAggregationDepthgetMaxBlockSizeInMBwritesave$init$, initializeForcefully, initializeLogIfNecessary, initializeLogIfNecessary, initializeLogIfNecessary$default$2, initLock, isTraceEnabled, log, logDebug, logDebug, logError, logError, logInfo, logInfo, logName, logTrace, logTrace, logWarning, logWarning, org$apache$spark$internal$Logging$$log__$eq, org$apache$spark$internal$Logging$$log_, uninitializepublic LinearRegression(String uid)
public LinearRegression()
public static LinearRegression load(String path)
public static int MAX_FEATURES_FOR_NORMAL_SOLVER()
LinearRegression.solver == "normal", the solver must limit the number of
features to at most this number. The entire covariance matrix X^T^X will be collected
to the driver. This limit helps prevent memory overflow errors.public static MLReader<T> read()
public final Param<String> solver()
LinearRegressionParamssolver in interface HasSolversolver in interface LinearRegressionParamspublic final Param<String> loss()
LinearRegressionParamsloss in interface HasLossloss in interface LinearRegressionParamspublic final DoubleParam epsilon()
LinearRegressionParamsepsilon in interface LinearRegressionParamspublic final DoubleParam maxBlockSizeInMB()
HasMaxBlockSizeInMBmaxBlockSizeInMB in interface HasMaxBlockSizeInMBpublic final IntParam aggregationDepth()
HasAggregationDepthaggregationDepth in interface HasAggregationDepthpublic final Param<String> weightCol()
HasWeightColweightCol in interface HasWeightColpublic final BooleanParam standardization()
HasStandardizationstandardization in interface HasStandardizationpublic final BooleanParam fitIntercept()
HasFitInterceptfitIntercept in interface HasFitInterceptpublic final DoubleParam tol()
HasTolpublic final IntParam maxIter()
HasMaxItermaxIter in interface HasMaxIterpublic final DoubleParam elasticNetParam()
HasElasticNetParamelasticNetParam in interface HasElasticNetParampublic final DoubleParam regParam()
HasRegParamregParam in interface HasRegParampublic String uid()
Identifiableuid in interface Identifiablepublic LinearRegression setRegParam(double value)
value - (undocumented)public LinearRegression setFitIntercept(boolean value)
value - (undocumented)public LinearRegression setStandardization(boolean value)
value - (undocumented)public LinearRegression setElasticNetParam(double value)
Note: Fitting with huber loss only supports None and L2 regularization, so throws exception if this param is non-zero value.
value - (undocumented)public LinearRegression setMaxIter(int value)
value - (undocumented)public LinearRegression setTol(double value)
value - (undocumented)public LinearRegression setWeightCol(String value)
value - (undocumented)public LinearRegression setSolver(String value)
LinearRegression.MAX_FEATURES_FOR_NORMAL_SOLVER.
- "auto" (default) means that the solver algorithm is selected automatically.
The Normal Equations solver will be used when possible, but this will automatically fall
back to iterative optimization methods when needed.
Note: Fitting with huber loss doesn't support normal solver, so throws exception if this param was set with "normal".
value - (undocumented)public LinearRegression setAggregationDepth(int value)
value - (undocumented)public LinearRegression setLoss(String value)
loss.
Default is "squaredError".
value - (undocumented)public LinearRegression setEpsilon(double value)
epsilon.
Default is 1.35.
value - (undocumented)public LinearRegression setMaxBlockSizeInMB(double value)
maxBlockSizeInMB.
Default is 0.0, then 1.0 MB will be chosen.
value - (undocumented)public LinearRegression copy(ParamMap extra)
ParamsdefaultCopy().copy in interface Paramscopy in class Predictor<Vector,LinearRegression,LinearRegressionModel>extra - (undocumented)