Package org.tribuo.classification.xgboost

Class XGBoostOptions

java.lang.Object

org.tribuo.classification.xgboost.XGBoostOptions

All Implemented Interfaces:

com.oracle.labs.mlrg.olcut.config.Options, ClassificationOptions<XGBoostClassificationTrainer>

public class XGBoostOptions
extends Object
implements ClassificationOptions<XGBoostClassificationTrainer>

CLI options for training an XGBoost classifier.

Field Summary

Fields

Modifier and Type	Field	Description
float	xbgAlpha	L1 regularization term for weights.
XGBoostTrainer.BoosterType	xgbBoosterType	Weak learning algorithm.
int	xgbEnsembleSize	Number of trees in the ensemble.
float	xgbEta	Step size shrinkage parameter (range [0,1]).
float	xgbGamma	Minimum loss reduction to make a split (range [0,Infinity]).
float	xgbLambda	L2 regularization term for weights.
XGBoostTrainer.LoggingVerbosity	xgbLogLevel	Make the XGBoost training procedure quiet.
int	xgbMaxDepth	Max tree depth (range (0,Integer.MAX_VALUE]).
float	xgbMinWeight	Minimum sum of instance weights needed in a leaf (range [0,Infinity]).
int	xgbNumThreads	Number of threads to use (range (1, num hw threads)).
boolean	xgbQuiet	Deprecated, use xgb-loglevel.
float	xgbSubsample	Subsample size for each tree (range (0,1]).
float	xgbSubsampleFeatures	Subsample features for each tree (range (0,1]).
XGBoostTrainer.TreeMethod	xgbTreeMethod	Tree building algorithm.

Fields inherited from interface com.oracle.labs.mlrg.olcut.config.Options

header

Constructor Summary

Constructors

Constructor	Description
XGBoostOptions()

Method Summary

Modifier and Type	Method	Description
XGBoostClassificationTrainer	getTrainer()	Constructs the trainer based on the provided arguments.

Methods inherited from class java.lang.Object

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

Methods inherited from interface com.oracle.labs.mlrg.olcut.config.Options

getOptionsDescription

Field Details

xgbBoosterType

@Option(longName="xgb-booster-type",
        usage="Weak learning algorithm.")
public XGBoostTrainer.BoosterType xgbBoosterType

Weak learning algorithm.

xgbTreeMethod

@Option(longName="xgb-tree-method",
        usage="Tree building algorithm.")
public XGBoostTrainer.TreeMethod xgbTreeMethod

Tree building algorithm.

xgbEnsembleSize

@Option(longName="xgb-ensemble-size",
        usage="Number of trees in the ensemble.")
public int xgbEnsembleSize

Number of trees in the ensemble.

xbgAlpha

@Option(longName="xgb-alpha",
        usage="L1 regularization term for weights.")
public float xbgAlpha

L1 regularization term for weights.

xgbMinWeight

@Option(longName="xgb-min-weight",
        usage="Minimum sum of instance weights needed in a leaf (range [0,Infinity]).")
public float xgbMinWeight

Minimum sum of instance weights needed in a leaf (range [0,Infinity]).

xgbMaxDepth

@Option(longName="xgb-max-depth",
        usage="Max tree depth (range (0,Integer.MAX_VALUE]).")
public int xgbMaxDepth

Max tree depth (range (0,Integer.MAX_VALUE]).

xgbEta

@Option(longName="xgb-eta",
        usage="Step size shrinkage parameter (range [0,1]).")
public float xgbEta

Step size shrinkage parameter (range [0,1]).

xgbSubsampleFeatures

@Option(longName="xgb-subsample-features",
        usage="Subsample features for each tree (range (0,1]).")
public float xgbSubsampleFeatures

Subsample features for each tree (range (0,1]).

xgbGamma

@Option(longName="xgb-gamma",
        usage="Minimum loss reduction to make a split (range [0,Infinity]).")
public float xgbGamma

Minimum loss reduction to make a split (range [0,Infinity]).

xgbLambda

@Option(longName="xgb-lambda",
        usage="L2 regularization term for weights.")
public float xgbLambda

L2 regularization term for weights.

xgbQuiet

@Option(longName="xgb-quiet",
        usage="Deprecated, use xgb-loglevel.")
public boolean xgbQuiet

Deprecated, use xgb-loglevel.

xgbLogLevel

@Option(longName="xgb-loglevel",
        usage="Make the XGBoost training procedure quiet.")
public XGBoostTrainer.LoggingVerbosity xgbLogLevel

Make the XGBoost training procedure quiet.

xgbSubsample

@Option(longName="xgb-subsample",
        usage="Subsample size for each tree (range (0,1]).")
public float xgbSubsample

Subsample size for each tree (range (0,1]).

xgbNumThreads

@Option(longName="xgb-num-threads",
        usage="Number of threads to use (range (1, num hw threads)). The default of 0 means use all hw threads.")
public int xgbNumThreads

Number of threads to use (range (1, num hw threads)). The default of 0 means use all hw threads.

Constructor Details

XGBoostOptions

public XGBoostOptions()

Method Details

getTrainer

public XGBoostClassificationTrainer getTrainer()

Description copied from interface: ClassificationOptions

Constructs the trainer based on the provided arguments.

Specified by:

getTrainer in interface ClassificationOptions<XGBoostClassificationTrainer>

Returns:

The trainer.