Package org.tribuo.regression.xgboost

Class XGBoostRegressionTrainer

java.lang.Object
org.tribuo.common.xgboost.XGBoostTrainer<Regressor>
org.tribuo.regression.xgboost.XGBoostRegressionTrainer
All Implemented Interfaces:
com.oracle.labs.mlrg.olcut.config.Configurable, com.oracle.labs.mlrg.olcut.provenance.Provenancable<TrainerProvenance>, Trainer<Regressor>, WeightedExamples
public final class XGBoostRegressionTrainer extends XGBoostTrainer<Regressor>
A Trainer which wraps the XGBoost training procedure. This only exposes a few of XGBoost's training parameters. It uses pthreads outside of the JVM to parallelise the computation.

Each output dimension is trained independently (and so contains a separate XGBoost ensemble).

See: 

 Chen T, Guestrin C.
 "XGBoost: A Scalable Tree Boosting System"
 Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016.
and for the original algorithm: 
 Friedman JH.
 "Greedy Function Approximation: a Gradient Boosting Machine"
 Annals of statistics, 2001.

N.B.: XGBoost4J wraps the native C implementation of xgboost that links to various C libraries, including libgomp and glibc (on Linux). If you're running on Alpine, which does not natively use glibc, you'll need to install glibc into the container. On the macOS binary on Maven Central is compiled without OpenMP support, meaning that XGBoost is single threaded on macOS. You can recompile the macOS binary with OpenMP support after installing libomp from homebrew if necessary.

  • Constructor Details

    • XGBoostRegressionTrainer

      public XGBoostRegressionTrainer(int numTrees)
      Creates an XGBoostRegressionTrainer using the default parameters, the squared error loss and the supplied number of trees.
      Parameters:
      numTrees - The number of trees.

    • XGBoostRegressionTrainer

      public XGBoostRegressionTrainer(XGBoostRegressionTrainer.RegressionType rType, int numTrees)
      Creates an XGBoostRegressionTrainer using the default parameters, the supplied loss and the supplied number of trees.
      Parameters:
      rType - The regression loss function.
      numTrees - The number of trees.

    • XGBoostRegressionTrainer

      public XGBoostRegressionTrainer(XGBoostRegressionTrainer.RegressionType rType, int numTrees, int numThreads, boolean silent)
      Creates an XGBoostRegressionTrainer using the default parameters with the supplied loss, number of trees, number of threads, and logging level.
      Parameters:
      rType - The regression loss function.
      numTrees - The number of trees.
      numThreads - The number of threads.
      silent - Silence the XGBoost logger.

    • XGBoostRegressionTrainer

      public XGBoostRegressionTrainer(XGBoostRegressionTrainer.RegressionType rType, int numTrees, double eta, double gamma, int maxDepth, double minChildWeight, double subsample, double featureSubsample, double lambda, double alpha, int nThread, boolean silent, long seed)
      Create an XGBoost trainer.
      Parameters:
      rType - The type of regression to build.
      numTrees - Number of trees to boost.
      eta - Step size shrinkage parameter (default 0.3, range [0,1]).
      gamma - Minimum loss reduction to make a split (default 0, range [0,inf]).
      maxDepth - Maximum tree depth (default 6, range [1,inf]).
      minChildWeight - Minimum sum of instance weights needed in a leaf (default 1, range [0, inf]).
      subsample - Subsample size for each tree (default 1, range (0,1]).
      featureSubsample - Subsample features for each tree (default 1, range (0,1]).
      lambda - L2 regularization term on weights (default 1).
      alpha - L1 regularization term on weights (default 0).
      nThread - Number of threads to use (default 4).
      silent - Silence the training output text.
      seed - RNG seed.

    • XGBoostRegressionTrainer

      public XGBoostRegressionTrainer(XGBoostTrainer.BoosterType boosterType, XGBoostTrainer.TreeMethod treeMethod, XGBoostRegressionTrainer.RegressionType rType, int numTrees, double eta, double gamma, int maxDepth, double minChildWeight, double subsample, double featureSubsample, double lambda, double alpha, int nThread, XGBoostTrainer.LoggingVerbosity verbosity, long seed)
      Create an XGBoost trainer.
      Parameters:
      boosterType - The base learning algorithm.
      treeMethod - The tree building algorithm if using a tree booster.
      rType - The type of regression to build.
      numTrees - Number of trees to boost.
      eta - Step size shrinkage parameter (default 0.3, range [0,1]).
      gamma - Minimum loss reduction to make a split (default 0, range [0,inf]).
      maxDepth - Maximum tree depth (default 6, range [1,inf]).
      minChildWeight - Minimum sum of instance weights needed in a leaf (default 1, range [0, inf]).
      subsample - Subsample size for each tree (default 1, range (0,1]).
      featureSubsample - Subsample features for each tree (default 1, range (0,1]).
      lambda - L2 regularization term on weights (default 1).
      alpha - L1 regularization term on weights (default 0).
      nThread - Number of threads to use (default 4).
      verbosity - Set the logging verbosity of the native library.
      seed - RNG seed.

    • XGBoostRegressionTrainer

      public XGBoostRegressionTrainer(XGBoostRegressionTrainer.RegressionType rType, int numTrees, Map<String,Object> parameters)
      This gives direct access to the XGBoost parameter map.

      It lets you pick things that we haven't exposed like dropout trees, binary classification etc.

      This sidesteps the validation that Tribuo provides for the hyperparameters, and so can produce unexpected results.

      Parameters:
      rType - The type of the regression.
      numTrees - Number of trees to boost.
      parameters - A map from string to object, where object can be Number or String.

  • Method Details

    • postConfig

      public void postConfig()
      Used by the OLCUT configuration system, and should not be called by external code.
      Specified by:
      postConfig in interface com.oracle.labs.mlrg.olcut.config.Configurable
      Overrides:
      postConfig in class XGBoostTrainer<Regressor>

    • train

      public XGBoostModel<Regressor> train(Dataset<Regressor> examples, Map<String,com.oracle.labs.mlrg.olcut.provenance.Provenance> runProvenance)
      Description copied from interface: Trainer
      Trains a predictive model using the examples in the given data set.
      Parameters:
      examples - the data set containing the examples.
      runProvenance - Training run specific provenance (e.g., fold number).
      Returns:
      a predictive model that can be used to generate predictions for new examples.

    • train

      public XGBoostModel<Regressor> train(Dataset<Regressor> examples, Map<String,com.oracle.labs.mlrg.olcut.provenance.Provenance> runProvenance, int invocationCount)
      Description copied from interface: Trainer
      Trains a predictive model using the examples in the given data set.
      Parameters:
      examples - the data set containing the examples.
      runProvenance - Training run specific provenance (e.g., fold number).
      invocationCount - The invocation counter that the trainer should be set to before training, which in most cases alters the state of the RNG inside this trainer. If the value is set to Trainer.INCREMENT_INVOCATION_COUNT then the invocation count is not changed.
      Returns:
      a predictive model that can be used to generate predictions for new examples.

    • getProvenance

      public TrainerProvenance getProvenance()