package params
Type Members
-
abstract
class
BaseTrainParams extends Serializable
Defines the common Booster parameters passed to the LightGBM learners.
-
case class
CategoricalParams(minDataPerGroup: Option[Int], maxCatThreshold: Option[Int], catl2: Option[Double], catSmooth: Option[Double], maxCatToOneHot: Option[Int]) extends ParamGroup with Product with Serializable
Defines parameters related to categorical features for lightgbm.
Defines parameters related to categorical features for lightgbm.
- minDataPerGroup
minimal number of data per categorical group.
- maxCatThreshold
limit number of split points considered for categorical features
- catl2
L2 regularization in categorical split.
- catSmooth
this can reduce the effect of noises in categorical features, especially for categories with few data.
- maxCatToOneHot
when number of categories of one feature smaller than or equal to this, one-vs-other split algorithm will be used.
-
case class
ClassifierTrainParams(passThroughArgs: Option[String], isUnbalance: Boolean, boostFromAverage: Boolean, isProvideTrainingMetric: Option[Boolean], delegate: Option[LightGBMDelegate], generalParams: GeneralParams, datasetParams: DatasetParams, dartModeParams: DartModeParams, executionParams: ExecutionParams, objectiveParams: ObjectiveParams, seedParams: SeedParams, categoricalParams: CategoricalParams, numClass: Int = 1) extends BaseTrainParams with Product with Serializable
Defines the Booster parameters passed to the LightGBM classifier.
-
case class
DartModeParams(dropRate: Double, maxDrop: Int, skipDrop: Double, xgboostDartMode: Boolean, uniformDrop: Boolean) extends ParamGroup with Product with Serializable
Defines the dart mode parameters passed to the LightGBM learners.
-
case class
DatasetParams(isEnableSparse: Option[Boolean], useMissing: Option[Boolean], zeroAsMissing: Option[Boolean]) extends ParamGroup with Product with Serializable
Defines the Dataset parameters passed to the LightGBM classifier.
Defines the Dataset parameters passed to the LightGBM classifier.
- isEnableSparse
Used to enable/disable sparse optimization.
- useMissing
Set this to false to disable the special handle of missing value.
- zeroAsMissing
Set to true to treat all zero as missing values (including the unshown values in LibSVM/sparse matrices) Set to false to use na for representing missing values.
-
case class
ExecutionParams(chunkSize: Int, matrixType: String, numThreads: Int, dataTransferMode: String, samplingMode: String, samplingSetSize: Int, microBatchSize: Int, useSingleDatasetMode: Boolean, maxStreamingOMPThreads: Int) extends ParamGroup with Product with Serializable
Defines parameters related to lightgbm execution in spark.
Defines parameters related to lightgbm execution in spark.
- chunkSize
Advanced parameter to specify the chunk size for copying Java data to native.
- matrixType
Advanced parameter to specify whether the native lightgbm matrix constructed should be sparse or dense.
- numThreads
The number of threads to run the native lightgbm training with on each worker.
- dataTransferMode
How to transfer data to LightGBM to begin the processing.
- samplingMode
How to sample data.
- samplingSetSize
The size of the subset if sampling only a subset.
- microBatchSize
The number of elements in a streaming micro-batch.
- useSingleDatasetMode
Whether to create only 1 LightGBM Dataset on each worker.
- maxStreamingOMPThreads
Maximum number of streaming mode OpenMP threads per Spark Task thread.
-
class
FObjParam extends ComplexParam[FObjTrait]
Param for FObjTrait.
Param for FObjTrait. Needed as spark has explicit params for many different types but not FObjTrait.
- trait FObjTrait extends Serializable
-
case class
GeneralParams(parallelism: String, topK: Option[Int], numIterations: Int, learningRate: Double, numLeaves: Option[Int], maxBin: Option[Int], binSampleCount: Option[Int], baggingFraction: Option[Double], posBaggingFraction: Option[Double], negBaggingFraction: Option[Double], baggingFreq: Option[Int], baggingSeed: Option[Int], earlyStoppingRound: Int, improvementTolerance: Double, featureFraction: Option[Double], featureFractionByNode: Option[Double], maxDepth: Option[Int], minSumHessianInLeaf: Option[Double], numMachines: Int, modelString: Option[String], categoricalFeatures: Array[Int], verbosity: Int, boostingType: String, lambdaL1: Option[Double], lambdaL2: Option[Double], metric: Option[String], minGainToSplit: Option[Double], maxDeltaStep: Option[Double], maxBinByFeature: Array[Int], minDataPerBin: Option[Int], minDataInLeaf: Option[Int], topRate: Option[Double], otherRate: Option[Double], monotoneConstraints: Array[Int], monotoneConstraintsMethod: Option[String], monotonePenalty: Option[Double], featureNames: Array[String]) extends ParamGroup with Product with Serializable
Defines the general Booster parameters passed to the LightGBM library.
-
trait
LightGBMBinParams extends Wrappable
Defines common parameters across all LightGBM learners related to histogram bin construction.
-
class
LightGBMBoosterParam extends ComplexParam[LightGBMBooster] with WrappableParam[LightGBMBooster]
Custom ComplexParam for LightGBMBooster, to make it settable on the LightGBM models.
-
trait
LightGBMCategoricalParams extends Wrappable
Defines common parameters across all LightGBM learners related to categorical variable treatment.
-
trait
LightGBMDartParams extends Wrappable
Defines parameters for dart mode across all LightGBM learners.
-
trait
LightGBMDatasetParams extends Wrappable
Defines common parameters across all LightGBM learners related to dataset handling.
-
trait
LightGBMExecutionParams extends Wrappable
Defines common LightGBM execution parameters.
-
trait
LightGBMFractionParams extends Wrappable
Defines parameters for fraction across all LightGBM learners.
-
trait
LightGBMLearnerParams extends Wrappable
Defines common parameters across all LightGBM learners related to learning score evolution.
-
trait
LightGBMModelParams extends Wrappable
Defines parameters for LightGBM models
-
trait
LightGBMObjectiveParams extends Wrappable
Defines common objective parameters
-
trait
LightGBMParams extends Wrappable with DefaultParamsWritable with HasWeightCol with HasValidationIndicatorCol with HasInitScoreCol with LightGBMExecutionParams with LightGBMSlotParams with LightGBMFractionParams with LightGBMBinParams with LightGBMLearnerParams with LightGBMDatasetParams with LightGBMDartParams with LightGBMPredictionParams with LightGBMObjectiveParams with LightGBMSeedParams with LightGBMCategoricalParams
Defines common parameters across all LightGBM learners.
-
trait
LightGBMPredictionParams extends Wrappable
Defines common prediction parameters across LightGBM Ranker, Classifier and Regressor
-
trait
LightGBMSeedParams extends Wrappable
Defines common parameters related to seed and determinism
-
trait
LightGBMSlotParams extends Wrappable
Defines parameters for slots across all LightGBM learners.
-
case class
ObjectiveParams(objective: String, fobj: Option[FObjTrait]) extends ParamGroup with Product with Serializable
Defines parameters related to the lightgbm objective function.
Defines parameters related to the lightgbm objective function.
- objective
The Objective. For regression applications, this can be: regression_l2, regression_l1, huber, fair, poisson, quantile, mape, gamma or tweedie. For classification applications, this can be: binary, multiclass, or multiclassova.
- fobj
Customized objective function. Should accept two parameters: preds, train_data, and return (grad, hess).
-
case class
RankerTrainParams(passThroughArgs: Option[String], maxPosition: Int, labelGain: Array[Double], evalAt: Array[Int], isProvideTrainingMetric: Option[Boolean], delegate: Option[LightGBMDelegate], generalParams: GeneralParams, datasetParams: DatasetParams, dartModeParams: DartModeParams, executionParams: ExecutionParams, objectiveParams: ObjectiveParams, seedParams: SeedParams, categoricalParams: CategoricalParams) extends BaseTrainParams with Product with Serializable
Defines the Booster parameters passed to the LightGBM ranker.
-
case class
RegressorTrainParams(passThroughArgs: Option[String], alpha: Double, tweedieVariancePower: Double, boostFromAverage: Boolean, isProvideTrainingMetric: Option[Boolean], delegate: Option[LightGBMDelegate], generalParams: GeneralParams, datasetParams: DatasetParams, dartModeParams: DartModeParams, executionParams: ExecutionParams, objectiveParams: ObjectiveParams, seedParams: SeedParams, categoricalParams: CategoricalParams) extends BaseTrainParams with Product with Serializable
Defines the Booster parameters passed to the LightGBM regressor.
-
case class
SeedParams(seed: Option[Int], deterministic: Option[Boolean], baggingSeed: Option[Int], featureFractionSeed: Option[Int], extraSeed: Option[Int], dropSeed: Option[Int], dataRandomSeed: Option[Int], objectiveSeed: Option[Int], boostingType: String, objective: String) extends ParamGroup with Product with Serializable
Defines parameters related to seed and determinism for lightgbm.
Defines parameters related to seed and determinism for lightgbm.
- seed
Main seed, used to generate other seeds.
- deterministic
Setting this to true should ensure stable results when using the same data and the same parameters.
- baggingSeed
Bagging seed.
- featureFractionSeed
Feature fraction seed.
- extraSeed
Random seed for selecting threshold when extra_trees is true.
- dropSeed
Random seed to choose dropping models. Only used in dart.
- dataRandomSeed
Random seed for sampling data to construct histogram bins.
- objectiveSeed
Random seed for objectives, if random process is needed. Currently used only for rank_xendcg objective.
- boostingType
Boosting type, used to determine if drop seed should be set.
- objective
Objective, used to determine if objective seed should be set.