API Guide

Article
10/13/2022

Note

Several listed components will be available in future releases and currently their doc reference links are disabled, for example, the TimeSeries transforms below (ExponentialAverage, SlidingWindow).

Trainers
Transforms
Subclasses
Loss Functions

Trainers

Click on link to see details for each of the trainers. Some classifiers implement predict_proba() to produce calibrated probabilities. Others implement decision_function() only for raw scores.

Binary Classifiers

Trainer	predict_proba()	decision_function()
`AveragedPerceptronBinaryClassifier`	Yes	Yes
`FactorizationMachineBinaryClassifier`	Yes	Yes
`FastForestBinaryClassifier`	Yes	Yes
`FastLinearBinaryClassifier`	Yes	Yes
`FastTreesBinaryClassifier`	Yes	Yes
`GamBinaryClassifier`	Yes	Yes
`LightGbmBinaryClassifier`	Yes	Yes
`LinearSvmBinaryClassifier`	Yes	Yes
`LocalDeepSvmBinaryClassifier`		Yes
`LogisticRegressionBinaryClassifier`	Yes	Yes
`SgdBinaryClassifier`	Yes	Yes
`SymSgdBinaryClassifier`	Yes	Yes

Multiclass Classifiers

Trainer	predict_proba()	decision_function()
`EnsembleClassifier`	Yes	Yes
`FastLinearClassifier`	Yes	Yes
`LightGbmClassifier`	Yes	Yes
`LogisticRegressionClassifier`	Yes
`NaiveBayesClassifier`		Yes
`OneVsRestClassifier`	Yes	Yes

Regressors

Trainer
`EnsembleRegressor`
`FastForestRegressor`
`FastLinearRegressor`
`FastTreesRegressor`
`FastTreesTweedieRegressor`
`GamRegressor`
`LightGbmRegressor`
`OnlineGradientDescentRegressor`
`OrdinaryLeastSquaresRegressor`
`PoissonRegressionRegressor`

Others

Trainer	Type
`LightGbmRanker`	ranker
`KMeansPlusPlus`	clusterer
`OneClassSvmAnomalyDetector`	anomaly
`PcaAnomalyDetector`	anomaly

Transforms

Click on link to see details for each of the transforms, and dependent subclasses.

Feature Extraction

Transform	Additional subclasses
`LightLda`
`Loader`
`NGramFeaturizer`	`Ngram`, `NgramHash`, `CustomStopWordsRemover`, `PredefinedStopWordsRemover`
`OneHotHashVectorizer`
`OneHotVectorizer`
`PcaTransformer`
`PixelExtractor`
`Resizer`
`Sentiment`
`TreeFeaturizer`
`WordEmbedding`

Feature Selection

Transform
`CountSelector`
`MutualInformationSelector`

Preprocessing

Transform
`Binner`
`BootstrapSampler`
`CharTokenizer`
`ColumnConcatenator`
`ColumnDropper`
`ColumnDuplicator`
`ColumnSelector`
`Expression`
`Filter`
`GlobalContrastRowScaler`
`Handler`
`Indicator`
`FromKey`
`LogMeanVarianceScaler`
`LpScaler`
`MeanVarianceScaler`
`MinMaxScaler`
`PrefixColumnConcatenator`
`RangeFilter`
`SkipFilter`
`SupervisedBinner`
`TakeFilter`
`TensorFlowScorer`
`ToKey`
`TypeConverter`
`WordTokenizer`

TimeSeries

Transform
`ExponentialAverage`
`IIDChangePointDetector`
`IIDSpikeDetector`
`PercentileThreshold`
`Pvalue`
`SlidingWindow`
`SsaChangePointDetector`
`SsaForecaster`
`SsaSpikeDetector`

Subclasses

These are auxillary classes used by transforms or trainers.

Subclasses	Used By
`AllFeatureSelector`	`AllInstanceSelector`, `BootstrapSelector`, `RandomPartitionSelector`
`AllInstanceSelector`	`EnsembleClassifier`, `EnsembleRegressor`
`BootstrapSelector`	`EnsembleClassifier`, `EnsembleRegressor`
`ClassifierAllSelector`	`EnsembleClassifier`
`ClassifierAverage`	`EnsembleClassifier`
`ClassifierBestDiverseSelector`	`EnsembleClassifier`
`ClassifierBestPerformanceSelector`	`EnsembleClassifier`
`ClassifierDisagreement`	`ClassifierBestDiverseSelector`
`ClassifierMedian`	`EnsembleClassifier`
`ClassifierStacking`	`EnsembleClassifier`
`ClassifierVoting`	`EnsembleClassifier`
`ClassifierWeightedAverage`	`EnsembleClassifier`
`CustomStopWordsRemover`	`NGramFeaturizer`
`Dart`	`LightGbmBinaryClassifier`, `LightGbmClassifier`, `LightGbmRanker`, `LightGbmRegressor`
`Gbdt`	`LightGbmBinaryClassifier`, `LightGbmClassifier`, `LightGbmRanker`, `LightGbmRegressor`
`Goss`	`LightGbmBinaryClassifier`, `LightGbmClassifier`, `LightGbmRanker`, `LightGbmRegressor`
`LinearKernel`	`OneClassSvmAnomalyDetector`
`Ngram`	`NGramFeaturizer`
`NgramHash`	`NGramFeaturizer`
`PolynomialKernel`	`OneClassSvmAnomalyDetector`
`PredefinedStopWordsRemover`	`NGramFeaturizer`
`RandomFeatureSelector`	`AllInstanceSelector`, `BootstrapSelector`, `RandomPartitionSelector`
`RandomPartitionSelector`	`EnsembleClassifier`, `EnsembleRegressor`
`RbfKernel`	`OneClassSvmAnomalyDetector`
`RegressorAllSelector`	`EnsembleRegressor`
`RegressorAverage`	`EnsembleRegressor`
`RegressorBestDiverseSelector`	`EnsembleRegressor`
`RegressorBestPerformanceSelector`	`EnsembleRegressor`
`RegressorDisagreement`	`RegressorBestDiverseSelector`
`RegressorMedian`	`EnsembleRegressor`
`RegressorStacking`	`EnsembleRegressor`
`SigmoidKernel`	`OneClassSvmAnomalyDetector`

Loss Functions

Trainers use a variety of loss functions. Click on the links for further details about each of these.

Loss Functions	Used By
`Exp`	`AveragedPerceptronBinaryClassifier`, `SgdBinaryClassifier`
`Hinge`	`AveragedPerceptronBinaryClassifier`, `SgdBinaryClassifier`, `FastLinearBinaryClassifier`, `FastLinearClassifier`
`Log`	`AveragedPerceptronBinaryClassifier`, `SgdBinaryClassifier`, `FastLinearBinaryClassifier`, `FastLinearClassifier`
`Poisson`	`OnlineGradientDescentRegressor`
`SmoothedHinge`	`AveragedPerceptronBinaryClassifier`, `SgdBinaryClassifier`, `FastLinearBinaryClassifier`, `FastLinearClassifier`
`Squared`	`FastLinearRegressor`, `OnlineGradientDescentRegressor`
`Tweedie`	`OnlineGradientDescentRegressor`