public static SweepablePipelineDataContract ToDataContract(this SweepablePipeline pipeline)
{
var estimatorContracts = new List <List <SweepableEstimatorDataContract> >();
var nodes = pipeline.EstimatorGenerators;
foreach (var node in nodes)
{
var estimators = new List <SweepableEstimatorDataContract>();
for (int i = 0; i != node.Count; ++i)
{
var estimator = node[i].RawValue as SweepableEstimatorBase;
var estimatorContract = new SweepableEstimatorDataContract()
{
EstimatorName = estimator.EstimatorName,
InputColumns = estimator.InputColumns,
OutputColumns = estimator.OutputColumns,
Scope = estimator.Scope,
};
estimators.Add(estimatorContract);
}
estimatorContracts.Add(estimators);
}
return(new SweepablePipelineDataContract()
{
Estimators = estimatorContracts,
});
}
public static SingleEstimatorSweepablePipelineDataContract ToDataContract(this SingleEstimatorSweepablePipeline pipeline)
{
var estimatorContracts = new List <SweepableEstimatorDataContract>();
var nodes = pipeline.Estimators;
foreach (var node in nodes)
{
var estimatorContract = new SweepableEstimatorDataContract()
{
EstimatorName = node.EstimatorName,
InputColumns = node.InputColumns,
OutputColumns = node.OutputColumns,
Scope = node.Scope,
};
estimatorContracts.Add(estimatorContract);
}
return(new SingleEstimatorSweepablePipelineDataContract()
{
Estimators = estimatorContracts,
});
}
public SweepableEstimatorBase CreateSweepableEstimator(SweepableEstimatorDataContract estimator)
{
var input = estimator.InputColumns[0];
var output = estimator.OutputColumns[0];
switch (estimator.EstimatorName)
{
case nameof(LightGbmRegressionTrainer):
var label = estimator.InputColumns[0];
var feature = estimator.InputColumns[1];
return(this.Context.AutoML().Regression.LightGbm(label, feature));
case nameof(LinearSvmTrainer):
label = estimator.InputColumns[0];
feature = estimator.InputColumns[1];
return(this.Context.AutoML().Serializable().BinaryClassification.LinearSvm(label, feature));
case nameof(LdSvmTrainer):
label = estimator.InputColumns[0];
feature = estimator.InputColumns[1];
return(this.Context.AutoML().Serializable().BinaryClassification.LdSvm(label, feature));
case nameof(FastForestBinaryTrainer):
label = estimator.InputColumns[0];
feature = estimator.InputColumns[1];
return(this.Context.AutoML().Serializable().BinaryClassification.FastForest(label, feature));
case nameof(FastTreeBinaryTrainer):
label = estimator.InputColumns[0];
feature = estimator.InputColumns[1];
return(this.Context.AutoML().Serializable().BinaryClassification.FastTree(label, feature));
case nameof(LightGbmBinaryTrainer):
label = estimator.InputColumns[0];
feature = estimator.InputColumns[1];
return(this.Context.AutoML().Serializable().BinaryClassification.LightGbm(label, feature));
case nameof(GamBinaryTrainer):
label = estimator.InputColumns[0];
feature = estimator.InputColumns[1];
return(this.Context.AutoML().Serializable().BinaryClassification.Gam(label, feature));
case nameof(SgdNonCalibratedTrainer):
label = estimator.InputColumns[0];
feature = estimator.InputColumns[1];
return(this.Context.AutoML().Serializable().BinaryClassification.SgdNonCalibrated(label, feature));
case nameof(SgdCalibratedTrainer):
label = estimator.InputColumns[0];
feature = estimator.InputColumns[1];
return(this.Context.AutoML().Serializable().BinaryClassification.SgdCalibrated(label, feature));
case nameof(AveragedPerceptronTrainer):
label = estimator.InputColumns[0];
feature = estimator.InputColumns[1];
return(this.Context.AutoML().Serializable().BinaryClassification.AveragedPerceptron(label, feature));
case nameof(OneHotEncodingEstimator):
return(this.Context.AutoML().Serializable().Transformer.Categorical.OneHotEncoding(input, output));
case nameof(MissingValueReplacingEstimator):
return(this.Context.AutoML().Serializable().Transformer.ReplaceMissingValues(input, output));
case nameof(ColumnConcatenatingEstimator):
return(this.Context.AutoML().Serializable().Transformer.Concatnate(estimator.InputColumns, output));
case nameof(TextFeaturizingEstimator):
return(this.Context.AutoML().Serializable().Transformer.Text.FeaturizeText(input, output));
case nameof(SerializableTextCatalog.FeaturizeTextWithWordEmbedding):
return(this.Context.AutoML().Serializable().Transformer.Text.FeaturizeTextWithWordEmbedding(input, output));
default:
throw new Exception($"{estimator.EstimatorName} can't be created through SweepabeEstimatorFactory");
}
}
