Microsoft.ML.Trainers.FastTree Namespace

Options for boosting tree trainers.

Consecutive Loss in Generality (UP).

Early stopping rule used to terminate training process once meeting a specified criterion. Used for setting EarlyStoppingRuleEarlyStoppingRule.

A IEstimator to transform input feature vector to tree-based features.

Options for the FastForestBinaryFeaturizationEstimator.

Model parameters for FastForestBinaryTrainer.

The IEstimator for training a decision tree binary classification model using Fast Forest.

Options for the FastForestBinaryTrainer as used in FastForest(Options).

Base class for fast forest trainer options.

A IEstimator to transform input feature vector to tree-based features.

Options for the FastForestRegressionFeaturizationEstimator.

Model parameters for FastForestRegressionTrainer.

The IEstimator for training a decision tree regression model using Fast Forest.

Options for the FastForestRegressionTrainer as used in FastForest(Options).

A IEstimator to transform input feature vector to tree-based features.

Options for the FastTreeBinaryFeaturizationEstimator.

Model parameters for FastTreeBinaryTrainer.

The IEstimator for training a decision tree binary classification model using FastTree.

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

A IEstimator to transform input feature vector to tree-based features.

Options for the FastTreeRankingFeaturizationEstimator.

Model parameters for FastTreeRankingTrainer.

The IEstimator for training a decision tree ranking model using FastTree.

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

A IEstimator to transform input feature vector to tree-based features.

Options for the FastTreeRegressionFeaturizationEstimator.

Model parameters for FastForestRegressionTrainer.

The IEstimator for training a decision tree regression model using FastTree.

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

A IEstimator to transform input feature vector to tree-based features.

Options for the FastTreeTweedieFeaturizationEstimator.

Model parameters for FastTreeTweedieTrainer.

The IEstimator for training a decision tree regression model using Tweedie loss function. This trainer is a generalization of Poisson, compound Poisson, and gamma regression.

Options for the FastTreeTweedieTrainer as used in FastTreeTweedie(Options).

Model parameters for GamBinaryTrainer.

The IEstimator for training a binary classification model with generalized additive models (GAM).

Options for the GamBinaryTrainer as used in Gam(Options).

The base class for GAM Model Parameters.

Model parameters for GamRegressionTrainer.

The IEstimator for training a regression model with generalized additive models (GAM).

Options for the GamRegressionTrainer as used in Gam(Options).

Base class for GAM-based trainer options.

Base class for GAM trainers.

Loss of Generality (GL).

Generality to Progress Ratio (PQ).

Low Progress (LP). This rule fires when the improvements on the score stall.

A IEstimator which contains a pre-trained TreeEnsembleModelParameters and calling its Fit(IDataView) produces a featurizer based on the pre-trained model.

PretrainedTreeFeaturizationEstimator.Options of PretrainedTreeFeaturizationEstimator as used when calling FeaturizeByPretrainTreeEnsemble(TransformsCatalog, PretrainedTreeFeaturizationEstimator+Options).

A container class for exposing Microsoft.ML.Trainers.FastTree.InternalQuantileRegressionTree's attributes to users. This class should not be mutable, so it contains a lot of read-only members. In addition to things inherited from RegressionTreeBase, we add GetLeafSamplesAt(Int32) and GetLeafSampleWeightsAt(Int32) to expose (sub-sampled) training labels falling into the leafIndex-th leaf and their weights.

A container class for exposing Microsoft.ML.Trainers.FastTree.InternalRegressionTree's attributes to users. This class should not be mutable, so it contains a lot of read-only members. Note that RegressionTree is identical to RegressionTreeBase but in another derived class QuantileRegressionTree some attributes are added.

A container base class for exposing Microsoft.ML.Trainers.FastTree.InternalRegressionTree's and Microsoft.ML.Trainers.FastTree.InternalQuantileRegressionTree's attributes to users. This class should not be mutable, so it contains a lot of read-only members.

A list of RegressionTreeBase's derived class. To compute the output value of a TreeEnsemble, we need to compute the output values of all trees in Trees, scale those values via TreeWeights, and finally sum the scaled values and Bias up.

This class encapsulates the common behavior of all tree-based featurizers such as FastTreeBinaryFeaturizationEstimator, FastForestBinaryFeaturizationEstimator, FastTreeRegressionFeaturizationEstimator, FastForestRegressionFeaturizationEstimator, and PretrainedTreeFeaturizationEstimator. All tree-based featurizers share the same output schema computed by GetOutputSchema(SchemaShape). All tree-based featurizers requires an input feature column name and a suffix for all output columns. The ITransformer returned by Fit(IDataView) produces three columns: (1) the prediction values of all trees, (2) the IDs of leaves the input feature vector falling into, and (3) the binary vector which encodes the paths to those destination leaves.

The common options of tree-based featurizations such as FastTreeBinaryFeaturizationEstimator, FastForestBinaryFeaturizationEstimator, FastTreeRegressionFeaturizationEstimator, FastForestRegressionFeaturizationEstimator, and PretrainedTreeFeaturizationEstimator.

ITransformer resulting from fitting any derived class of TreeEnsembleFeaturizationEstimatorBase. The derived classes include, for example, FastTreeBinaryFeaturizationEstimator and FastForestRegressionFeaturizationEstimator.

TreeEnsembleModelParametersBasedOnQuantileRegressionTree is derived from TreeEnsembleModelParameters plus a strongly-typed public attribute, TrainedTreeEnsemble, for exposing trained model's details to users. Its function, Microsoft.ML.Trainers.FastTree.TreeEnsembleModelParametersBasedOnQuantileRegressionTree.CreateTreeEnsembleFromInternalDataStructure, is called to create TrainedTreeEnsemble inside TreeEnsembleModelParameters. Note that the major difference between TreeEnsembleModelParametersBasedOnQuantileRegressionTree and TreeEnsembleModelParametersBasedOnRegressionTree is the type of TrainedTreeEnsemble.

TreeEnsembleModelParametersBasedOnRegressionTree is derived from TreeEnsembleModelParameters plus a strongly-typed public attribute, TrainedTreeEnsemble, for exposing trained model's details to users. Its function, Microsoft.ML.Trainers.FastTree.TreeEnsembleModelParametersBasedOnRegressionTree.CreateTreeEnsembleFromInternalDataStructure, is called to create TrainedTreeEnsemble inside TreeEnsembleModelParameters. Note that the major difference between TreeEnsembleModelParametersBasedOnQuantileRegressionTree and TreeEnsembleModelParametersBasedOnRegressionTree is the type of TrainedTreeEnsemble.

Options for tree trainers.

Types of optimization algorithms.

Stopping measurements for classification and regression.

Stopping measurements for ranking.