hex.tree

Class SharedTree<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

java.lang.Object

water.Iced

hex.ModelBuilder<M,P,O>

hex.tree.SharedTree<M,P,O>

All Implemented Interfaces:

CalibrationHelper.ModelBuilderWithCalibration<M,P,O>, java.io.Externalizable, java.io.Serializable, java.lang.Cloneable, water.Freezable

Direct Known Subclasses:

DRF, GBM, IsolationForest, UpliftDRF

public abstract class SharedTree<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>
extends hex.ModelBuilder<M,P,O>
implements CalibrationHelper.ModelBuilderWithCalibration<M,P,O>

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
protected class	SharedTree.Driver
protected static class	SharedTree.FrameMap
protected static class	SharedTree.ScoreBuildOneTree
static class	SharedTree.SharedTreeDebugParams

Nested classes/interfaces inherited from class hex.ModelBuilder

hex.ModelBuilder.BuilderVisibility, hex.ModelBuilder.FilterCols, hex.ModelBuilder.ModelTrainingCoordinator, hex.ModelBuilder.ValidationMessage

Field Summary

Fields

Modifier and Type	Field and Description
protected GlobalInteractionConstraints	_ics
protected float[]	_improvPerVar
protected double	_initialPrediction
protected int	_lastScoredTree
protected M	_model
protected int	_mtry
protected int	_mtry_per_tree
protected int	_ncols
protected int	_ntrees
protected java.util.Random	_rand
protected water.fvec.Frame	_trainPredsCache
protected water.fvec.Frame	_validPredsCache
protected water.fvec.Frame	_validWorkspace
static int	MAX_NTREES

Fields inherited from class hex.ModelBuilder

_coordinator, _desc, _eventPublisher, _fold, _input_parms, _job, _messages, _nclass, _offset, _orig_projection_array, _origDomains, _origNames, _origTrain, _parms, _priorClassDist, _removedCols, _response, _result, _startUpOnceModelBuilder, _train, _treatment, _valid, _vresponse, _weights

Constructor Summary

Constructors

Constructor and Description
SharedTree(P parms)
SharedTree(P parms, boolean startup_once)
SharedTree(P parms, water.Job job)
SharedTree(P parms, water.Key<M> key)

Method Summary

Modifier and Type	Method and Description
protected void	addCustomInfo(O out)
protected DHistogram[][][]	buildLayer(water.fvec.Frame fr, int nbins, DTree[] ktrees, int[] leafs, DHistogram[][][] hcs, boolean build_tree_one_node)
protected boolean	canLearnFromNAs()
protected void	checkEarlyStoppingReproducibility()
protected void	checkMemoryFootPrint_impl()
protected water.fvec.Chunk	chk_nids(water.fvec.Chunk[] chks, int c)
protected water.fvec.Chunk	chk_offset(water.fvec.Chunk[] chks)
protected water.fvec.Chunk	chk_oobt(water.fvec.Chunk[] chks)
water.fvec.Chunk	chk_resp(water.fvec.Chunk[] chks)
water.fvec.Chunk	chk_tree(water.fvec.Chunk[] chks, int c)
water.fvec.Chunk	chk_weight(water.fvec.Chunk[] chks)
protected water.fvec.Chunk	chk_work(water.fvec.Chunk[] chks, int c)
protected boolean	computePriorClassDistribution()
static water.util.TwoDimTable	createModelSummaryTable(int ntrees, TreeStats treeStats)
protected water.util.TwoDimTable	createScoringHistoryTable()
static water.util.TwoDimTable	createScoringHistoryTable(hex.Model.Output _output, hex.ScoreKeeper[] _scored_train, hex.ScoreKeeper[] _scored_valid, water.Job job, long[] _training_time_ms, boolean hasCustomMetric, boolean hasCustomDistribution)
protected boolean	cv_canBuildMainModelInParallel()
void	cv_computeAndSetOptimalParameters(hex.ModelBuilder<M,P,O>[] cvModelBuilders)
protected boolean	cv_initStoppingParameters()
protected boolean	cv_updateOptimalParameters(hex.ModelBuilder<M,P,O>[] cvModelBuilders)
protected double[]	data_row(water.fvec.Chunk[] chks, int row, double[] data)
protected boolean	doScoringAndSaveModel(boolean finalScoring, boolean oob, boolean build_tree_one_node)
water.fvec.Frame	getCalibrationFrame()
protected double	getInitialValue() Compute the inital value for a given distribution
hex.ModelBuilder	getModelBuilder()
protected hex.ScoreKeeper.ProblemType	getProblemType()
hex.ToEigenVec	getToEigenVec()
boolean	haveMojo()
boolean	havePojo()
protected int	idx_nids(int c)
protected int	idx_offset()
protected int	idx_oobt()
protected int	idx_resp()
protected int	idx_treatment()
protected int	idx_tree(int c)
protected int	idx_weight()
protected int	idx_work(int c)
protected void	ignoreInvalidColumns(int npredictors, boolean expensive)
void	init(boolean expensive) Initialize the ModelBuilder, validating all arguments and preparing the training frame.
boolean	isSupervised()
boolean	isUplift()
protected DTree.DecidedNode	makeDecided(DTree.UndecidedNode udn, DHistogram[] hs, Constraints cs)
protected Score.ScoreExtension	makeScoreExtension()
boolean	providesVarImp()
protected abstract double	score1(water.fvec.Chunk[] chks, double offset, double weight, double[] fs, int row)
boolean	scoreZeroTrees()
void	setCalibrationFrame(water.fvec.Frame f)
void	setDebugParams(SharedTree.SharedTreeDebugParams debugParms) Modify algorithm inner workings - only meant for development
boolean	shouldReorder(water.fvec.Vec v)
protected void	validateRowSampleRate()
protected water.fvec.Frame	validWorkspace()
protected water.fvec.Vec	vec_nids(water.fvec.Frame fr, int c)
protected water.fvec.Vec	vec_offset(water.fvec.Frame fr)
protected water.fvec.Vec	vec_oobt(water.fvec.Frame fr)
protected water.fvec.Vec	vec_resp(water.fvec.Frame fr)
protected water.fvec.Vec	vec_tree(water.fvec.Frame fr, int c)
protected water.fvec.Vec	vec_weight(water.fvec.Frame fr)
protected water.fvec.Vec	vec_work(water.fvec.Frame fr, int c)

Methods inherited from class hex.ModelBuilder

algoName, algos, builderVisibility, can_build, checkCustomMetricForEarlyStopping, checkDistributions, checkMemoryFootPrint, checkResponseVariable, clearInitState, clearValidationErrors, computeCrossValidation, cv_buildModels, cv_mainModelScores, cv_makeAggregateModelMetrics, cv_scoreCVModels, defaultKey, desiredChunks, dest, error_count, error, get, getMessagesByFieldAndSeverity, getName, getSysProperty, hasFoldCol, hasOffsetCol, hasTreatmentCol, hasWeightCol, hide, ignoreBadColumns, ignoreConstColumns, ignoreStringColumns, ignoreUuidColumns, info, init_adaptFrameToTrain, init_getNClass, initWorkspace, isClassifier, isResponseOptional, isStopped, javaName, logMe, make, make, make, makeCVMetrics, makeCVModelBuilder, makeParameters, makePojoWriter, message, nclasses, nFoldCV, nFoldWork, nModelsInParallel, nModelsInParallel, nModelsInParallel, numSpecialCols, paramName, raiseReproducibilityWarning, rebalance, remainingTimeSecs, response, schemaDirectory, separateFeatureVecs, setMaxRuntimeSecsForMainModel, setTrain, setValid, smallDataSize, stop_requested, timeout, train, trainModel, trainModel, trainModelImpl, trainModelNested, trainModelNested, trainModelOnH2ONode, valid, validateBinaryResponse, validateStoppingMetric, validationErrors, validationWarnings, vresponse, warn

Methods inherited from class water.Iced

asBytes, clone, copyOver, frozenType, read, readExternal, readJSON, reloadFromBytes, toJsonBytes, toJsonString, write, writeExternal, writeJSON

Methods inherited from class java.lang.Object

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

Field Detail

_mtry

protected int _mtry

_mtry_per_tree

protected int _mtry_per_tree

_ics

protected GlobalInteractionConstraints _ics

MAX_NTREES

public static final int MAX_NTREES

See Also:

Constant Field Values

_ntrees

protected int _ntrees

_model

protected M extends SharedTreeModel<M,P,O> _model

_ncols

protected int _ncols

_initialPrediction

protected double _initialPrediction

_improvPerVar

protected transient float[] _improvPerVar

_rand

protected java.util.Random _rand

_validWorkspace

protected transient water.fvec.Frame _validWorkspace

_lastScoredTree

protected transient int _lastScoredTree

_trainPredsCache

protected transient water.fvec.Frame _trainPredsCache

_validPredsCache

protected transient water.fvec.Frame _validPredsCache

Constructor Detail

SharedTree

public SharedTree(P parms)

SharedTree

public SharedTree(P parms,
                  water.Key<M> key)

SharedTree

public SharedTree(P parms,
                  water.Job job)

SharedTree

public SharedTree(P parms,
                  boolean startup_once)

Method Detail

shouldReorder

public boolean shouldReorder(water.fvec.Vec v)

Overrides:

shouldReorder in class hex.ModelBuilder<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

validWorkspace

protected final water.fvec.Frame validWorkspace()

isSupervised

public boolean isSupervised()

Specified by:

isSupervised in class hex.ModelBuilder<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

isUplift

public boolean isUplift()

providesVarImp

public boolean providesVarImp()

makeScoreExtension

protected Score.ScoreExtension makeScoreExtension()

haveMojo

public boolean haveMojo()

Overrides:

haveMojo in class hex.ModelBuilder<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

havePojo

public boolean havePojo()

Overrides:

havePojo in class hex.ModelBuilder<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

scoreZeroTrees

public boolean scoreZeroTrees()

computePriorClassDistribution

protected boolean computePriorClassDistribution()

Overrides:

computePriorClassDistribution in class hex.ModelBuilder<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

getToEigenVec

public hex.ToEigenVec getToEigenVec()

Overrides:

getToEigenVec in class hex.ModelBuilder<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

ignoreInvalidColumns

protected void ignoreInvalidColumns(int npredictors,
                                    boolean expensive)

Overrides:

ignoreInvalidColumns in class hex.ModelBuilder<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

init

public void init(boolean expensive)

Initialize the ModelBuilder, validating all arguments and preparing the training frame. This call is expected to be overridden in the subclasses and each subclass will start with "super.init();". This call is made by the front-end whenever the GUI is clicked, and needs to be fast; heavy-weight prep needs to wait for the trainModel() call. Validate the requested ntrees; precompute actual ntrees. Validate the number of classes to predict on; validate a checkpoint.

Overrides:

init in class hex.ModelBuilder<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

validateRowSampleRate

protected void validateRowSampleRate()

checkEarlyStoppingReproducibility

protected void checkEarlyStoppingReproducibility()

Overrides:

checkEarlyStoppingReproducibility in class hex.ModelBuilder<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

getProblemType

protected hex.ScoreKeeper.ProblemType getProblemType()

buildLayer

protected DHistogram[][][] buildLayer(water.fvec.Frame fr,
                                      int nbins,
                                      DTree[] ktrees,
                                      int[] leafs,
                                      DHistogram[][][] hcs,
                                      boolean build_tree_one_node)

idx_weight

protected int idx_weight()

idx_offset

protected int idx_offset()

idx_resp

protected int idx_resp()

idx_tree

protected int idx_tree(int c)

idx_work

protected int idx_work(int c)

idx_nids

protected int idx_nids(int c)

idx_oobt

protected int idx_oobt()

idx_treatment

protected int idx_treatment()

chk_weight

public water.fvec.Chunk chk_weight(water.fvec.Chunk[] chks)

chk_offset

protected water.fvec.Chunk chk_offset(water.fvec.Chunk[] chks)

chk_resp

public water.fvec.Chunk chk_resp(water.fvec.Chunk[] chks)

chk_tree

public water.fvec.Chunk chk_tree(water.fvec.Chunk[] chks,
                                 int c)

chk_work

protected water.fvec.Chunk chk_work(water.fvec.Chunk[] chks,
                                    int c)

chk_nids

protected water.fvec.Chunk chk_nids(water.fvec.Chunk[] chks,
                                    int c)

chk_oobt

protected water.fvec.Chunk chk_oobt(water.fvec.Chunk[] chks)

vec_weight

protected final water.fvec.Vec vec_weight(water.fvec.Frame fr)

vec_offset

protected final water.fvec.Vec vec_offset(water.fvec.Frame fr)

vec_resp

protected final water.fvec.Vec vec_resp(water.fvec.Frame fr)

vec_tree

protected final water.fvec.Vec vec_tree(water.fvec.Frame fr,
                                        int c)

vec_work

protected final water.fvec.Vec vec_work(water.fvec.Frame fr,
                                        int c)

vec_nids

protected final water.fvec.Vec vec_nids(water.fvec.Frame fr,
                                        int c)

vec_oobt

protected final water.fvec.Vec vec_oobt(water.fvec.Frame fr)

data_row

protected double[] data_row(water.fvec.Chunk[] chks,
                            int row,
                            double[] data)

makeDecided

protected DTree.DecidedNode makeDecided(DTree.UndecidedNode udn,
                                        DHistogram[] hs,
                                        Constraints cs)

score1

protected abstract double score1(water.fvec.Chunk[] chks,
                                 double offset,
                                 double weight,
                                 double[] fs,
                                 int row)

doScoringAndSaveModel

protected final boolean doScoringAndSaveModel(boolean finalScoring,
                                              boolean oob,
                                              boolean build_tree_one_node)

getModelBuilder

public hex.ModelBuilder getModelBuilder()

Specified by:

getModelBuilder in interface CalibrationHelper.ModelBuilderWithCalibration<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

getCalibrationFrame

public final water.fvec.Frame getCalibrationFrame()

Specified by:

getCalibrationFrame in interface CalibrationHelper.ModelBuilderWithCalibration<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

setCalibrationFrame

public void setCalibrationFrame(water.fvec.Frame f)

Specified by:

setCalibrationFrame in interface CalibrationHelper.ModelBuilderWithCalibration<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

canLearnFromNAs

protected boolean canLearnFromNAs()

Overrides:

canLearnFromNAs in class hex.ModelBuilder<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

addCustomInfo

protected void addCustomInfo(O out)

createScoringHistoryTable

protected water.util.TwoDimTable createScoringHistoryTable()

createScoringHistoryTable

public static water.util.TwoDimTable createScoringHistoryTable(hex.Model.Output _output,
                                                               hex.ScoreKeeper[] _scored_train,
                                                               hex.ScoreKeeper[] _scored_valid,
                                                               water.Job job,
                                                               long[] _training_time_ms,
                                                               boolean hasCustomMetric,
                                                               boolean hasCustomDistribution)

createModelSummaryTable

public static water.util.TwoDimTable createModelSummaryTable(int ntrees,
                                                             TreeStats treeStats)

checkMemoryFootPrint_impl

protected void checkMemoryFootPrint_impl()

Overrides:

checkMemoryFootPrint_impl in class hex.ModelBuilder<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

getInitialValue

protected double getInitialValue()

Compute the inital value for a given distribution

Returns:

initial value

cv_canBuildMainModelInParallel

protected boolean cv_canBuildMainModelInParallel()

Overrides:

cv_canBuildMainModelInParallel in class hex.ModelBuilder<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

cv_computeAndSetOptimalParameters

public void cv_computeAndSetOptimalParameters(hex.ModelBuilder<M,P,O>[] cvModelBuilders)

Overrides:

cv_computeAndSetOptimalParameters in class hex.ModelBuilder<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

cv_updateOptimalParameters

protected final boolean cv_updateOptimalParameters(hex.ModelBuilder<M,P,O>[] cvModelBuilders)

Overrides:

cv_updateOptimalParameters in class hex.ModelBuilder<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

cv_initStoppingParameters

protected final boolean cv_initStoppingParameters()

Overrides:

cv_initStoppingParameters in class hex.ModelBuilder<M extends SharedTreeModel<M,P,O>,P extends SharedTreeModel.SharedTreeParameters,O extends SharedTreeModel.SharedTreeOutput>

setDebugParams

public void setDebugParams(SharedTree.SharedTreeDebugParams debugParms)

Modify algorithm inner workings - only meant for development

Parameters:

debugParms - instance of SharedTreeDebugParams