FastTreeRegressionTrainer.Options Class

Definition

Options for the FastTreeRegressionTrainer as used in FastTree(Options).

public sealed class FastTreeRegressionTrainer.Options : Microsoft.ML.Trainers.FastTree.BoostedTreeOptions, Microsoft.ML.Runtime.IComponentFactory<Microsoft.ML.ITrainer>
type FastTreeRegressionTrainer.Options = class
    inherit BoostedTreeOptions
    interface IComponentFactory<ITrainer>
    interface IComponentFactory
Public NotInheritable Class FastTreeRegressionTrainer.Options
Inherits BoostedTreeOptions
Implements IComponentFactory(Of ITrainer)
Inheritance
Implements
IComponentFactory IComponentFactory<Microsoft.ML.ITrainer>

Constructors

FastTreeRegressionTrainer.Options()

Create a new FastTreeRegressionTrainer.Options object with default values.

Fields

AllowEmptyTrees

When a root split is impossible, allow training to proceed.

(Inherited from TreeOptions)
BaggingExampleFraction

Percentage of training examples used in each bag. Default is 0.7 (70%).

(Inherited from TreeOptions)
BaggingSize

Number of trees in each bag (0 for disabling bagging).

(Inherited from TreeOptions)
BestStepRankingRegressionTrees

Option for using best regression step trees.

(Inherited from BoostedTreeOptions)
Bias

Bias for calculating gradient for each feature bin for a categorical feature.

(Inherited from TreeOptions)
Bundling

Bundle low population bins. Bundle.None(0): no bundling, Bundle.AggregateLowPopulation(1): Bundle low population, Bundle.Adjacent(2): Neighbor low population bundle.

(Inherited from TreeOptions)
CategoricalSplit

Whether to do split based on multiple categorical feature values.

(Inherited from TreeOptions)
CompressEnsemble

Compress the tree Ensemble.

(Inherited from TreeOptions)
DiskTranspose

Whether to utilize the disk or the data's native transposition facilities (where applicable) when performing the transpose.

(Inherited from TreeOptions)
DropoutRate

Dropout rate for tree regularization.

(Inherited from BoostedTreeOptions)
EnablePruning

Enable post-training tree pruning to avoid overfitting. It requires a validation set.

(Inherited from BoostedTreeOptions)
EntropyCoefficient

The entropy (regularization) coefficient between 0 and 1.

(Inherited from TreeOptions)
ExampleWeightColumnName

Column to use for example weight.

(Inherited from TrainerInputBaseWithWeight)
ExecutionTime

Print execution time breakdown to ML.NET channel.

(Inherited from TreeOptions)
FeatureColumnName

Column to use for features.

(Inherited from TrainerInputBase)
FeatureFirstUsePenalty

The feature first use penalty coefficient.

(Inherited from TreeOptions)
FeatureFlocks

Whether to collectivize features during dataset preparation to speed up training.

(Inherited from TreeOptions)
FeatureFraction

The fraction of features (chosen randomly) to use on each iteration. Use 0.9 if only 90% of features is needed. Lower numbers help reduce over-fitting.

(Inherited from TreeOptions)
FeatureFractionPerSplit

The fraction of features (chosen randomly) to use on each split. If it's value is 0.9, 90% of all features would be dropped in expectation.

(Inherited from TreeOptions)
FeatureReusePenalty

The feature re-use penalty (regularization) coefficient.

(Inherited from TreeOptions)
FeatureSelectionSeed

The seed of the active feature selection.

(Inherited from TreeOptions)
FilterZeroLambdas

Filter zero lambdas during training.

(Inherited from BoostedTreeOptions)
GainConfidenceLevel

Tree fitting gain confidence requirement. Only consider a gain if its likelihood versus a random choice gain is above this value.

(Inherited from TreeOptions)
GetDerivativesSampleRate

Sample each query 1 in k times in the GetDerivatives function.

(Inherited from BoostedTreeOptions)
HistogramPoolSize

The number of histograms in the pool (between 2 and numLeaves).

(Inherited from TreeOptions)
LabelColumnName

Column to use for labels.

(Inherited from TrainerInputBaseWithLabel)
LearningRate

The learning rate.

(Inherited from BoostedTreeOptions)
MaximumBinCountPerFeature

Maximum number of distinct values (bins) per feature.

(Inherited from TreeOptions)
MaximumCategoricalGroupCountPerNode

Maximum categorical split groups to consider when splitting on a categorical feature. Split groups are a collection of split points. This is used to reduce overfitting when there many categorical features.

(Inherited from TreeOptions)
MaximumCategoricalSplitPointCount

Maximum categorical split points to consider when splitting on a categorical feature.

(Inherited from TreeOptions)
MaximumNumberOfLineSearchSteps

Number of post-bracket line search steps.

(Inherited from BoostedTreeOptions)
MaximumTreeOutput

Upper bound on absolute value of single tree output.

(Inherited from BoostedTreeOptions)
MemoryStatistics

Print memory statistics to ML.NET channel.

(Inherited from TreeOptions)
MinimumExampleCountPerLeaf

The minimal number of data points required to form a new tree leaf.

(Inherited from TreeOptions)
MinimumExampleFractionForCategoricalSplit

Minimum categorical example percentage in a bin to consider for a split. Default is 0.1% of all training examples.

(Inherited from TreeOptions)
MinimumExamplesForCategoricalSplit

Minimum categorical example count in a bin to consider for a split.

(Inherited from TreeOptions)
MinimumStepSize

Minimum line search step size.

(Inherited from BoostedTreeOptions)
NumberOfLeaves

The max number of leaves in each regression tree.

(Inherited from TreeOptions)
NumberOfThreads

The number of threads to use.

(Inherited from TreeOptions)
NumberOfTrees

Total number of decision trees to create in the ensemble.

(Inherited from TreeOptions)
OptimizationAlgorithm

Optimization algorithm to be used.

(Inherited from BoostedTreeOptions)
PruningThreshold

The tolerance threshold for pruning.

(Inherited from BoostedTreeOptions)
PruningWindowSize

The moving window size for pruning.

(Inherited from BoostedTreeOptions)
RandomStart

Training starts from random ordering (determined by /r1).

(Inherited from BoostedTreeOptions)
RowGroupColumnName

Column to use for example groupId.

(Inherited from TrainerInputBaseWithGroupId)
Seed

The seed of the random number generator.

(Inherited from TreeOptions)
Shrinkage

Shrinkage.

(Inherited from BoostedTreeOptions)
Smoothing

Smoothing parameter for tree regularization.

(Inherited from TreeOptions)
SoftmaxTemperature

The temperature of the randomized softmax distribution for choosing the feature.

(Inherited from TreeOptions)
SparsifyThreshold

Sparsity level needed to use sparse feature representation.

(Inherited from TreeOptions)
TestFrequency

Calculate metric values for train/valid/test every k rounds.

(Inherited from TreeOptions)
UseLineSearch

Determines whether to use line search for a step size.

(Inherited from BoostedTreeOptions)
UseTolerantPruning

Use window and tolerance for pruning.

(Inherited from BoostedTreeOptions)
WriteLastEnsemble

Write the last ensemble instead of the one determined by early stopping.

(Inherited from BoostedTreeOptions)

Properties

EarlyStoppingMetric

Early stopping metrics.

EarlyStoppingRule

Early stopping rule used to terminate training process once meeting a specified criterion. Possible choices are EarlyStoppingRuleBase's implementations such as TolerantEarlyStoppingRule and GeneralityLossRule.

(Inherited from BoostedTreeOptions)

Explicit Interface Implementations

IComponentFactory<ITrainer>.CreateComponent(IHostEnvironment)

Applies to