/// <summary>
/// Predict a target using a linear classification model trained with the <see cref="SgdNonCalibratedTrainer"/> trainer.
/// </summary>
/// <param name="catalog">The binary classification catalog trainer object.</param>
/// <param name="label">The name of the label column.</param>
/// <param name="features">The name of the feature column.</param>
/// <param name="weights">The name for the example weight column.</param>
/// <param name="numberOfIterations">The maximum number of iterations; set to 1 to simulate online learning.</param>
/// <param name="learningRate">The initial learning rate used by SGD.</param>
/// <param name="l2Regularization">The L2 weight for <a href='https://en.wikipedia.org/wiki/Regularization_(mathematics)'>regularization</a>.</param>
/// <param name="lossFunction">The loss function to use.</param>
/// <param name="onFit">A delegate that is called every time the
/// <see cref="Estimator{TTupleInShape, TTupleOutShape, TTransformer}.Fit(DataView{TTupleInShape})"/> method is called on the
/// <see cref="Estimator{TTupleInShape, TTupleOutShape, TTransformer}"/> instance created out of this. This delegate will receive
/// the linear model that was trained. Note that this action cannot change the result in any way; it is only a way for the caller to
/// be informed about what was learnt.</param>
/// <returns>The predicted output.</returns>
public static (Scalar <float> score, Scalar <bool> predictedLabel) StochasticGradientDescentNonCalibratedClassificationTrainer(
this BinaryClassificationCatalog.BinaryClassificationTrainers catalog,
Scalar <bool> label,
Vector <float> features,
Scalar <float> weights = null,
int numberOfIterations = SgdNonCalibratedTrainer.Options.Defaults.NumberOfIterations,
double learningRate = SgdNonCalibratedTrainer.Options.Defaults.LearningRate,
float l2Regularization = SgdNonCalibratedTrainer.Options.Defaults.L2Regularization,
IClassificationLoss lossFunction = null,
Action <LinearBinaryModelParameters> onFit = null)
{
var rec = new TrainerEstimatorReconciler.BinaryClassifierNoCalibration(
(env, labelName, featuresName, weightsName) =>
{
var trainer = new SgdNonCalibratedTrainer(env, labelName, featuresName, weightsName,
numberOfIterations, learningRate, l2Regularization, lossFunction);
if (onFit != null)
{
return(trainer.WithOnFitDelegate(trans => onFit(trans.Model)));
}
return(trainer);
}, label, features, weights);
return(rec.Output);
}
/// <summary>
/// Predict a target using a linear binary classification model trained with the AveragedPerceptron trainer, and a custom loss.
/// </summary>
/// <param name="catalog">The binary classification catalog trainer object.</param>
/// <param name="label">The label, or dependent variable.</param>
/// <param name="features">The features, or independent variables.</param>
/// <param name="lossFunction">The custom loss.</param>
/// <param name="weights">The optional example weights.</param>
/// <param name="learningRate">The learning Rate.</param>
/// <param name="decreaseLearningRate">Decrease learning rate as iterations progress.</param>
/// <param name="l2Regularization">L2 regularization weight.</param>
/// <param name="numIterations">Number of training iterations through the data.</param>
/// <param name="onFit">A delegate that is called every time the
/// <see cref="Estimator{TInShape, TOutShape, TTransformer}.Fit(DataView{TInShape})"/> method is called on the
/// <see cref="Estimator{TInShape, TOutShape, TTransformer}"/> instance created out of this. This delegate will receive
/// the linear model that was trained, as well as the calibrator on top of that model. Note that this action cannot change the
/// result in any way; it is only a way for the caller to be informed about what was learnt.</param>
/// <returns>The set of output columns including in order the predicted binary classification score (which will range
/// from negative to positive infinity), and the predicted label.</returns>
/// <seealso cref="AveragedPerceptronTrainer"/>.
/// <example>
/// <format type="text/markdown">
/// <]
/// ]]></format>
/// </example>
public static (Scalar <float> score, Scalar <bool> predictedLabel) AveragedPerceptron(
this BinaryClassificationCatalog.BinaryClassificationTrainers catalog,
Scalar <bool> label,
Vector <float> features,
Scalar <float> weights = null,
IClassificationLoss lossFunction = null,
float learningRate = AveragedLinearOptions.AveragedDefault.LearningRate,
bool decreaseLearningRate = AveragedLinearOptions.AveragedDefault.DecreaseLearningRate,
float l2Regularization = AveragedLinearOptions.AveragedDefault.L2Regularization,
int numIterations = AveragedLinearOptions.AveragedDefault.NumberOfIterations,
Action <LinearBinaryModelParameters> onFit = null
)
{
OnlineLinearStaticUtils.CheckUserParams(label, features, weights, learningRate, l2Regularization, numIterations, onFit);
bool hasProbs = lossFunction is LogLoss;
var rec = new TrainerEstimatorReconciler.BinaryClassifierNoCalibration(
(env, labelName, featuresName, weightsName) =>
{
var trainer = new AveragedPerceptronTrainer(env, labelName, featuresName, lossFunction,
learningRate, decreaseLearningRate, l2Regularization, numIterations);
if (onFit != null)
{
return(trainer.WithOnFitDelegate(trans => onFit(trans.Model)));
}
else
{
return(trainer);
}
}, label, features, weights);
return(rec.Output);
}
/// <summary>
/// Predict a target using a linear binary classification model trained with the AveragedPerceptron trainer, and a custom loss.
/// </summary>
/// <param name="ctx">The binary classification context trainer object.</param>
/// <param name="label">The label, or dependent variable.</param>
/// <param name="features">The features, or independent variables.</param>
/// <param name="lossFunction">The custom loss.</param>
/// <param name="weights">The optional example weights.</param>
/// <param name="learningRate">The learning Rate.</param>
/// <param name="decreaseLearningRate">Decrease learning rate as iterations progress.</param>
/// <param name="l2RegularizerWeight">L2 regularization weight.</param>
/// <param name="numIterations">Number of training iterations through the data.</param>
/// <param name="onFit">A delegate that is called every time the
/// <see cref="Estimator{TTupleInShape, TTupleOutShape, TTransformer}.Fit(DataView{TTupleInShape})"/> method is called on the
/// <see cref="Estimator{TTupleInShape, TTupleOutShape, TTransformer}"/> instance created out of this. This delegate will receive
/// the linear model that was trained, as well as the calibrator on top of that model. Note that this action cannot change the
/// result in any way; it is only a way for the caller to be informed about what was learnt.</param>
/// <returns>The set of output columns including in order the predicted binary classification score (which will range
/// from negative to positive infinity), and the predicted label.</returns>
/// <seealso cref="AveragedPerceptronTrainer"/>.
public static (Scalar <float> score, Scalar <bool> predictedLabel) AveragedPerceptron(
this BinaryClassificationContext.BinaryClassificationTrainers ctx,
IClassificationLoss lossFunction,
Scalar <bool> label, Vector <float> features, Scalar <float> weights = null,
float learningRate = AveragedLinearArguments.AveragedDefaultArgs.LearningRate,
bool decreaseLearningRate = AveragedLinearArguments.AveragedDefaultArgs.DecreaseLearningRate,
float l2RegularizerWeight = AveragedLinearArguments.AveragedDefaultArgs.L2RegularizerWeight,
int numIterations = AveragedLinearArguments.AveragedDefaultArgs.NumIterations,
Action <LinearBinaryPredictor> onFit = null
)
{
OnlineLinearStaticUtils.CheckUserParams(label, features, weights, learningRate, l2RegularizerWeight, numIterations, onFit);
bool hasProbs = lossFunction is HingeLoss;
var args = new AveragedPerceptronTrainer.Arguments()
{
LearningRate = learningRate,
DecreaseLearningRate = decreaseLearningRate,
L2RegularizerWeight = l2RegularizerWeight,
NumIterations = numIterations
};
if (lossFunction != null)
{
args.LossFunction = new TrivialClassificationLossFactory(lossFunction);
}
var rec = new TrainerEstimatorReconciler.BinaryClassifierNoCalibration(
(env, labelName, featuresName, weightsName) =>
{
args.FeatureColumn = featuresName;
args.LabelColumn = labelName;
args.InitialWeights = weightsName;
var trainer = new AveragedPerceptronTrainer(env, args);
if (onFit != null)
{
return(trainer.WithOnFitDelegate(trans => onFit(trans.Model)));
}
else
{
return(trainer);
}
}, label, features, weights, hasProbs);
return(rec.Output);
}
/// <summary>
/// Predict a target using a linear binary classification model trained with <see cref="AveragedPerceptronTrainer"/>.
/// </summary>
/// <param name="catalog">The binary classification catalog trainer object.</param>
/// <param name="labelColumn">The name of the label column, or dependent variable.</param>
/// <param name="featureColumn">The features, or independent variables.</param>
/// <param name="lossFunction">A custom <a href="tmpurl_loss">loss</a>. If <see langword="null"/>, hinge loss will be used resulting in max-margin averaged perceptron.</param>
/// <param name="learningRate"><a href="tmpurl_lr">Learning rate</a>.</param>
/// <param name="decreaseLearningRate">
/// <see langword="true" /> to decrease the <paramref name="learningRate"/> as iterations progress; otherwise, <see langword="false" />.
/// Default is <see langword="false" />.
/// </param>
/// <param name="l2RegularizerWeight">L2 weight for <a href='tmpurl_regularization'>regularization</a>.</param>
/// <param name="numIterations">Number of passes through the training dataset.</param>
/// <example>
/// <format type="text/markdown">
/// <]
/// ]]>
/// </format>
/// </example>
public static AveragedPerceptronTrainer AveragedPerceptron(
this BinaryClassificationCatalog.BinaryClassificationTrainers catalog,
string labelColumn = DefaultColumnNames.Label,
string featureColumn = DefaultColumnNames.Features,
IClassificationLoss lossFunction = null,
float learningRate = AveragedLinearArguments.AveragedDefaultArgs.LearningRate,
bool decreaseLearningRate = AveragedLinearArguments.AveragedDefaultArgs.DecreaseLearningRate,
float l2RegularizerWeight = AveragedLinearArguments.AveragedDefaultArgs.L2RegularizerWeight,
int numIterations = AveragedLinearArguments.AveragedDefaultArgs.NumIterations)
{
Contracts.CheckValue(catalog, nameof(catalog));
var env = CatalogUtils.GetEnvironment(catalog);
return(new AveragedPerceptronTrainer(env, labelColumn, featureColumn, lossFunction ?? new LogLoss(), learningRate, decreaseLearningRate, l2RegularizerWeight, numIterations));
}
/// <summary>
/// Predict a target using a linear binary classification model trained with the AveragedPerceptron trainer, and a custom loss.
/// </summary>
/// <param name="ctx">The binary classification context trainer object.</param>
/// <param name="label">The label, or dependent variable.</param>
/// <param name="features">The features, or independent variables.</param>
/// <param name="lossFunction">The custom loss.</param>
/// <param name="weights">The optional example weights.</param>
/// <param name="learningRate">The learning Rate.</param>
/// <param name="decreaseLearningRate">Decrease learning rate as iterations progress.</param>
/// <param name="l2RegularizerWeight">L2 regularization weight.</param>
/// <param name="numIterations">Number of training iterations through the data.</param>
/// <param name="advancedSettings">A delegate to supply more advanced arguments to the algorithm.</param>
public static AveragedPerceptronTrainer AveragedPerceptron(
this BinaryClassificationContext.BinaryClassificationTrainers ctx,
string label = DefaultColumnNames.Label,
string features = DefaultColumnNames.Features,
string weights = null,
IClassificationLoss lossFunction = null,
float learningRate = AveragedLinearArguments.AveragedDefaultArgs.LearningRate,
bool decreaseLearningRate = AveragedLinearArguments.AveragedDefaultArgs.DecreaseLearningRate,
float l2RegularizerWeight = AveragedLinearArguments.AveragedDefaultArgs.L2RegularizerWeight,
int numIterations = AveragedLinearArguments.AveragedDefaultArgs.NumIterations,
Action <AveragedPerceptronTrainer.Arguments> advancedSettings = null)
{
Contracts.CheckValue(ctx, nameof(ctx));
var env = CatalogUtils.GetEnvironment(ctx);
return(new AveragedPerceptronTrainer(env, label, features, weights, lossFunction ?? new LogLoss(), learningRate, decreaseLearningRate, l2RegularizerWeight, numIterations, advancedSettings));
}
/// <summary>
/// Trains a linear binary classifier using the averaged perceptron.
/// <a href='https://en.wikipedia.org/wiki/Perceptron'>Wikipedia entry for Perceptron</a>
/// </summary>
/// <param name="env">The local instance of the <see cref="IHostEnvironment"/></param>
/// <param name="lossFunction">The classification loss function. </param>
/// <param name="labelColumnName">The name of the label column. </param>
/// <param name="featureColumnName">The name of the feature column.</param>
/// <param name="learningRate">The learning rate. </param>
/// <param name="decreaseLearningRate">Whether to decrease learning rate as iterations progress.</param>
/// <param name="l2Regularization">Weight of L2 regularization term.</param>
/// <param name="numberOfIterations">The number of training iterations.</param>
internal AveragedPerceptronTrainer(IHostEnvironment env,
string labelColumnName = DefaultColumnNames.Label,
string featureColumnName = DefaultColumnNames.Features,
IClassificationLoss lossFunction = null,
float learningRate = Options.AveragedDefault.LearningRate,
bool decreaseLearningRate = Options.AveragedDefault.DecreaseLearningRate,
float l2Regularization = Options.AveragedDefault.L2Regularization,
int numberOfIterations = Options.AveragedDefault.NumberOfIterations)
: this(env, new Options
{
LabelColumnName = labelColumnName,
FeatureColumnName = featureColumnName,
LearningRate = learningRate,
DecreaseLearningRate = decreaseLearningRate,
L2Regularization = l2Regularization,
NumberOfIterations = numberOfIterations,
LossFunction = lossFunction ?? new HingeLoss()
})
{
}
/// <summary>
/// Trains a linear binary classifier using the averaged perceptron.
/// <a href='https://en.wikipedia.org/wiki/Perceptron'>Wikipedia entry for Perceptron</a>
/// </summary>
/// <param name="env">The local instance of the <see cref="IHostEnvironment"/></param>
/// <param name="lossFunction">The classification loss function. </param>
/// <param name="labelColumn">The name of the label column. </param>
/// <param name="featureColumn">The name of the feature column.</param>
/// <param name="learningRate">The learning rate. </param>
/// <param name="decreaseLearningRate">Wheather to decrease learning rate as iterations progress.</param>
/// <param name="l2RegularizerWeight">L2 Regularization Weight.</param>
/// <param name="numIterations">The number of training iteraitons.</param>
internal AveragedPerceptronTrainer(IHostEnvironment env,
string labelColumn = DefaultColumnNames.Label,
string featureColumn = DefaultColumnNames.Features,
IClassificationLoss lossFunction = null,
float learningRate = Options.AveragedDefaultArgs.LearningRate,
bool decreaseLearningRate = Options.AveragedDefaultArgs.DecreaseLearningRate,
float l2RegularizerWeight = Options.AveragedDefaultArgs.L2RegularizerWeight,
int numIterations = Options.AveragedDefaultArgs.NumIterations)
: this(env, new Options
{
LabelColumn = labelColumn,
FeatureColumn = featureColumn,
LearningRate = learningRate,
DecreaseLearningRate = decreaseLearningRate,
L2RegularizerWeight = l2RegularizerWeight,
NumIterations = numIterations,
LossFunction = new TrivialFactory(lossFunction ?? new HingeLoss())
})
{
}
/// <summary>
/// Predict a target using a linear binary classification model trained with the AveragedPerceptron trainer, and a custom loss.
/// </summary>
/// <param name="catalog">The binary classification catalog trainer object.</param>
/// <param name="label">The label, or dependent variable.</param>
/// <param name="features">The features, or independent variables.</param>
/// <param name="lossFunction">The custom loss.</param>
/// <param name="weights">The optional example weights.</param>
/// <param name="options">Advanced arguments to the algorithm.</param>
/// <param name="onFit">A delegate that is called every time the
/// <see cref="Estimator{TInShape, TOutShape, TTransformer}.Fit(DataView{TInShape})"/> method is called on the
/// <see cref="Estimator{TInShape, TOutShape, TTransformer}"/> instance created out of this. This delegate will receive
/// the linear model that was trained, as well as the calibrator on top of that model. Note that this action cannot change the
/// result in any way; it is only a way for the caller to be informed about what was learnt.</param>
/// <returns>The set of output columns including in order the predicted binary classification score (which will range
/// from negative to positive infinity), and the predicted label.</returns>
/// <seealso cref="AveragedPerceptronTrainer"/>.
/// <example>
/// <format type="text/markdown">
/// <]
/// ]]></format>
/// </example>
public static (Scalar <float> score, Scalar <bool> predictedLabel) AveragedPerceptron(
this BinaryClassificationCatalog.BinaryClassificationTrainers catalog,
Scalar <bool> label,
Vector <float> features,
Scalar <float> weights,
IClassificationLoss lossFunction,
AveragedPerceptronTrainer.Options options,
Action <LinearBinaryModelParameters> onFit = null
)
{
Contracts.CheckValue(label, nameof(label));
Contracts.CheckValue(features, nameof(features));
Contracts.CheckValueOrNull(weights);
Contracts.CheckValueOrNull(options);
Contracts.CheckValueOrNull(onFit);
bool hasProbs = lossFunction is LogLoss;
var rec = new TrainerEstimatorReconciler.BinaryClassifierNoCalibration(
(env, labelName, featuresName, weightsName) =>
{
options.LabelColumn = labelName;
options.FeatureColumn = featuresName;
options.InitialWeights = weightsName;
var trainer = new AveragedPerceptronTrainer(env, options);
if (onFit != null)
{
return(trainer.WithOnFitDelegate(trans => onFit(trans.Model)));
}
else
{
return(trainer);
}
}, label, features, weights, hasProbs);
return(rec.Output);
}
public static PredictionEngine <TIn, TOut> AveragedPerceptron <TIn, TOut>(
IEnumerable <TIn> trainDataset,
IClassificationLoss lossFunction = null,
float learningRate = 1f,
bool decreaseLearningRate = false,
float l2Regularization = 0f,
int numberOfIterations = 1,
Action <ITransformer> additionModelAction = null)
where TIn : class, new()
where TOut : class, new()
{
var context = new MLContext();
var type = typeof(TIn);
var labelColumnName = Preprocessing.LabelColumn(type.GetProperties()).Name;
var properties = Preprocessing.ExcludeColumns(type.GetProperties());
var preprocessor = context.OneHotEncoding(properties);
var trainDataframe = context.Data.LoadFromEnumerable(trainDataset);
var pipeline = context.Transforms.Concatenate("Features", preprocessor.CombinedFeatures.ToArray())
.Append(preprocessor.OneHotEncodingEstimator)
.AppendCacheCheckpoint(context)
.Append(context.BinaryClassification.Trainers.AveragedPerceptron(
labelColumnName,
featureColumnName: "Features",
lossFunction,
learningRate,
decreaseLearningRate,
l2Regularization,
numberOfIterations
));
var model = pipeline.Fit(trainDataframe);
var predictEngine = context.Model.CreatePredictionEngine <TIn, TOut>(model);
additionModelAction?.Invoke(model);
return(predictEngine);
}
/// <summary>
/// Trains a linear binary classifier using the averaged perceptron.
/// <a href='https://en.wikipedia.org/wiki/Perceptron'>Wikipedia entry for Perceptron</a>
/// </summary>
/// <param name="env">The local instance of the <see cref="IHostEnvironment"/></param>
/// <param name="lossFunction">The classification loss function. </param>
/// <param name="labelColumn">The name of the label column. </param>
/// <param name="featureColumn">The name of the feature column.</param>
/// <param name="weights">The optional name of the weights column.</param>
/// <param name="learningRate">The learning rate. </param>
/// <param name="decreaseLearningRate">Wheather to decrease learning rate as iterations progress.</param>
/// <param name="l2RegularizerWeight">L2 Regularization Weight.</param>
/// <param name="numIterations">The number of training iteraitons.</param>
/// <param name="advancedSettings">A delegate to supply more advanced arguments to the algorithm.</param>
public AveragedPerceptronTrainer(IHostEnvironment env,
string labelColumn = DefaultColumnNames.Label,
string featureColumn = DefaultColumnNames.Features,
string weights = null,
IClassificationLoss lossFunction = null,
float learningRate = Arguments.AveragedDefaultArgs.LearningRate,
bool decreaseLearningRate = Arguments.AveragedDefaultArgs.DecreaseLearningRate,
float l2RegularizerWeight = Arguments.AveragedDefaultArgs.L2RegularizerWeight,
int numIterations = Arguments.AveragedDefaultArgs.NumIterations,
Action <Arguments> advancedSettings = null)
: this(env, InvokeAdvanced(advancedSettings, new Arguments
{
LabelColumn = labelColumn,
FeatureColumn = featureColumn,
InitialWeights = weights,
LearningRate = learningRate,
DecreaseLearningRate = decreaseLearningRate,
L2RegularizerWeight = l2RegularizerWeight,
NumIterations = numIterations,
LossFunction = new TrivialFactory(lossFunction ?? new HingeLoss())
}))
{
}
public TrivialClassificationLossFactory(IClassificationLoss loss)
{
_loss = loss;
}
/// <summary>
/// Predict a target using a linear classification model trained with the <see cref="SgdNonCalibratedBinaryTrainer"/> trainer.
/// </summary>
/// <param name="catalog">The binary classificaiton catalog trainer object.</param>
/// <param name="labelColumnName">The name of the label column.</param>
/// <param name="featureColumnName">The name of the feature column.</param>
/// <param name="weightColumnName">The name for the example weight column.</param>
/// <param name="loss">The loss function minimized in the training process. Using, for example, <see cref="HingeLoss"/> leads to a support vector machine trainer.</param>
/// <param name="maxIterations">The maximum number of iterations; set to 1 to simulate online learning.</param>
/// <param name="initLearningRate">The initial learning rate used by SGD.</param>
/// <param name="l2Weight">The L2 regularization constant.</param>
public static SgdNonCalibratedBinaryTrainer StochasticGradientDescentNonCalibrated(this BinaryClassificationCatalog.BinaryClassificationTrainers catalog,
string labelColumnName = DefaultColumnNames.Label,
string featureColumnName = DefaultColumnNames.Features,
string weightColumnName = null,
IClassificationLoss loss = null,
int maxIterations = SgdNonCalibratedBinaryTrainer.Options.Defaults.MaxIterations,
double initLearningRate = SgdNonCalibratedBinaryTrainer.Options.Defaults.InitLearningRate,
float l2Weight = SgdNonCalibratedBinaryTrainer.Options.Defaults.L2Weight)
{
Contracts.CheckValue(catalog, nameof(catalog));
var env = CatalogUtils.GetEnvironment(catalog);
return(new SgdNonCalibratedBinaryTrainer(env, labelColumnName, featureColumnName, weightColumnName,
maxIterations, initLearningRate, l2Weight, loss));
}
