public void OVAWithAllConstructorArgs()
{
var(pipeline, data) = GetMultiClassPipeline();
var calibrator = new PlattCalibratorTrainer(Env);
var averagePerceptron = new AveragedPerceptronTrainer(Env, "Label", "Features", advancedSettings: s =>
{
s.Shuffle = true;
s.Calibrator = null;
});
pipeline = pipeline.Append(new Ova(Env, averagePerceptron, "Label", true, calibrator: calibrator, 10000, true))
.Append(new KeyToValueMappingEstimator(Env, "PredictedLabel"));
TestEstimatorCore(pipeline, data);
Done();
}
public void OnlineLinearWorkout()
{
var dataPath = GetDataPath("breast-cancer.txt");
var regressionData = TextLoaderStatic.CreateReader(ML, ctx => (Label: ctx.LoadFloat(0), Features: ctx.LoadFloat(1, 10)))
.Read(dataPath);
var regressionPipe = regressionData.MakeNewEstimator()
.Append(r => (r.Label, Features: r.Features.Normalize()));
var regressionTrainData = regressionPipe.Fit(regressionData).Transform(regressionData).AsDynamic;
var ogdTrainer = new OnlineGradientDescentTrainer(ML, "Label", "Features");
TestEstimatorCore(ogdTrainer, regressionTrainData);
var ogdModel = ogdTrainer.Fit(regressionTrainData);
ogdTrainer.Train(regressionTrainData, ogdModel.Model);
var binaryData = TextLoaderStatic.CreateReader(ML, ctx => (Label: ctx.LoadBool(0), Features: ctx.LoadFloat(1, 10)))
.Read(dataPath);
var binaryPipe = binaryData.MakeNewEstimator()
.Append(r => (r.Label, Features: r.Features.Normalize()));
var binaryTrainData = binaryPipe.Fit(binaryData).Transform(binaryData).AsDynamic;
var apTrainer = new AveragedPerceptronTrainer(ML, "Label", "Features", lossFunction: new HingeLoss(), advancedSettings: s =>
{
s.LearningRate = 0.5f;
});
TestEstimatorCore(apTrainer, binaryTrainData);
var apModel = apTrainer.Fit(binaryTrainData);
apTrainer.Train(binaryTrainData, apModel.Model);
var svmTrainer = new LinearSvmTrainer(ML, "Label", "Features");
TestEstimatorCore(svmTrainer, binaryTrainData);
var svmModel = svmTrainer.Fit(binaryTrainData);
svmTrainer.Train(binaryTrainData, apModel.Model);
Done();
}
CalibratorTestData GetCalibratorTestData()
{
var(pipeline, data) = GetBinaryClassificationPipeline();
var binaryTrainer = new AveragedPerceptronTrainer(Env);
pipeline = pipeline.Append(binaryTrainer);
var transformer = pipeline.Fit(data);
var scoredData = transformer.Transform(data);
var scoredDataPreview = scoredData.Preview();
Assert.True(scoredDataPreview.ColumnView.Length == 5);
return(new CalibratorTestData
{
data = data,
scoredData = scoredData,
pipeline = pipeline,
transformer = ((TransformerChain <BinaryPredictionTransformer <LinearBinaryModelParameters> >)transformer).LastTransformer as BinaryPredictionTransformer <LinearBinaryModelParameters>,
});
}
public void OnlineLinearWorkout()
{
var dataPath = GetDataPath("breast-cancer.txt");
var data = TextLoader.CreateReader(Env, ctx => (Label: ctx.LoadFloat(0), Features: ctx.LoadFloat(1, 10)))
.Read(new MultiFileSource(dataPath));
var pipe = data.MakeNewEstimator()
.Append(r => (r.Label, Features: r.Features.Normalize()));
var trainData = pipe.Fit(data).Transform(data).AsDynamic;
IEstimator <ITransformer> est = new OnlineGradientDescentTrainer(Env, new OnlineGradientDescentTrainer.Arguments());
TestEstimatorCore(est, trainData);
est = new AveragedPerceptronTrainer(Env, new AveragedPerceptronTrainer.Arguments());
TestEstimatorCore(est, trainData);
Done();
}
/// <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 void OVAWithAllConstructorArgs()
{
var dataPath = GetDataPath(IrisDataPath);
string featNam = "Features";
string labNam = "Label";
using (var env = new TlcEnvironment())
{
var calibrator = new FixedPlattCalibratorTrainer(env, new FixedPlattCalibratorTrainer.Arguments());
var data = new TextLoader(env, GetIrisLoaderArgs()).Read(new MultiFileSource(dataPath));
var averagePerceptron = new AveragedPerceptronTrainer(env, new AveragedPerceptronTrainer.Arguments {
FeatureColumn = featNam, LabelColumn = labNam, Shuffle = true, Calibrator = null
});
var pipeline = new TermEstimator(env, labNam)
.Append(new Ova(env, averagePerceptron, labNam, true, calibrator: calibrator, 10000, true))
.Append(new KeyToValueEstimator(env, "PredictedLabel"));
TestEstimatorCore(pipeline, data);
}
}
public void New_TrainWithInitialPredictor()
{
using (var env = new LocalEnvironment(seed: 1, conc: 1))
{
var data = new TextLoader(env, MakeSentimentTextLoaderArgs()).Read(new MultiFileSource(GetDataPath(TestDatasets.Sentiment.trainFilename)));
// Pipeline.
var pipeline = new TextTransform(env, "SentimentText", "Features");
// Train the pipeline, prepare train set.
var trainData = pipeline.FitAndTransform(data);
// Train the first predictor.
var trainer = new LinearClassificationTrainer(env, "Features", "Label", advancedSettings: (s) => s.NumThreads = 1);
var firstModel = trainer.Fit(trainData);
// Train the second predictor on the same data.
var secondTrainer = new AveragedPerceptronTrainer(env, "Label", "Features");
var trainRoles = new RoleMappedData(trainData, label: "Label", feature: "Features");
var finalModel = secondTrainer.Train(new TrainContext(trainRoles, initialPredictor: firstModel.Model));
}
}
public void TrainWithInitialPredictor()
{
using (var env = new LocalEnvironment(seed: 1, conc: 1))
{
// Pipeline
var loader = TextLoader.ReadFile(env, MakeSentimentTextLoaderArgs(), new MultiFileSource(GetDataPath(TestDatasets.Sentiment.trainFilename)));
var trans = TextTransform.Create(env, MakeSentimentTextTransformArgs(), loader);
var trainData = trans;
var cachedTrain = new CacheDataView(env, trainData, prefetch: null);
// Train the first predictor.
var trainer = new LinearClassificationTrainer(env, new LinearClassificationTrainer.Arguments
{
NumThreads = 1
});
var trainRoles = new RoleMappedData(cachedTrain, label: "Label", feature: "Features");
var predictor = trainer.Train(new Runtime.TrainContext(trainRoles));
// Train the second predictor on the same data.
var secondTrainer = new AveragedPerceptronTrainer(env, "Label", "Features");
var finalPredictor = secondTrainer.Train(new TrainContext(trainRoles, initialPredictor: predictor));
}
}
public void OnlineLinearWorkout()
{
var dataPath = GetDataPath("breast-cancer.txt");
var data = TextLoader.CreateReader(Env, ctx => (Label: ctx.LoadFloat(0), Features: ctx.LoadFloat(1, 10)))
.Read(dataPath);
var pipe = data.MakeNewEstimator()
.Append(r => (r.Label, Features: r.Features.Normalize()));
var trainData = pipe.Fit(data).Transform(data).AsDynamic;
IEstimator <ITransformer> est = new OnlineGradientDescentTrainer(Env, "Label", "Features");
TestEstimatorCore(est, trainData);
est = new AveragedPerceptronTrainer(Env, "Label", "Features", lossFunction: new HingeLoss.Arguments(), advancedSettings: s =>
{
s.LearningRate = 0.5f;
});
TestEstimatorCore(est, trainData);
Done();
}
/// <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 avdanced 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 BinaryClassificationContext.BinaryClassificationTrainers ctx,
Scalar <bool> label,
Vector <float> features,
Scalar <float> 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,
Action <LinearBinaryModelParameters> onFit = null
)
{
OnlineLinearStaticUtils.CheckUserParams(label, features, weights, learningRate, l2RegularizerWeight, numIterations, onFit, advancedSettings);
bool hasProbs = lossFunction is LogLoss;
var rec = new TrainerEstimatorReconciler.BinaryClassifierNoCalibration(
(env, labelName, featuresName, weightsName) =>
{
var trainer = new AveragedPerceptronTrainer(env, labelName, featuresName, weightsName, lossFunction,
learningRate, decreaseLearningRate, l2RegularizerWeight, numIterations, advancedSettings);
if (onFit != null)
{
return(trainer.WithOnFitDelegate(trans => onFit(trans.Model)));
}
else
{
return(trainer);
}
}, label, features, weights, hasProbs);
return(rec.Output);
}
public MyAveragedPerceptron(IHostEnvironment env, AveragedPerceptronTrainer.Arguments args, string featureCol, string labelCol)
: base(env, new TrainerInfo(caching: false), featureCol, labelCol)
{
_trainer = new AveragedPerceptronTrainer(env, args);
}
private void MixMatch(string dataPath)
{
// Create a new environment for ML.NET operations. It can be used for exception tracking and logging,
// as well as the source of randomness.
var env = new LocalEnvironment();
// Read the data as an IDataView.
// First, we define the reader: specify the data columns and where to find them in the text file.
var reader = TextLoader.CreateReader(env, ctx => (
// The four features of the Iris dataset.
SepalLength: ctx.LoadFloat(0),
SepalWidth: ctx.LoadFloat(1),
PetalLength: ctx.LoadFloat(2),
PetalWidth: ctx.LoadFloat(3),
// Label: kind of iris.
Label: ctx.LoadText(4)
),
// Default separator is tab, but the dataset has comma.
separator: ',');
// Read the data.
var data = reader.Read(new MultiFileSource(dataPath));
// Build the pre-processing pipeline.
var learningPipeline = reader.MakeNewEstimator()
.Append(r => (
// Convert string label to a key.
Label: r.Label.ToKey(),
// Concatenate all the features together into one column 'Features'.
Features: r.SepalLength.ConcatWith(r.SepalWidth, r.PetalLength, r.PetalWidth)));
// Now, at the time of writing, there is no static pipeline for OVA (one-versus-all). So, let's
// append the OVA learner to the dynamic pipeline.
IEstimator <ITransformer> dynamicPipe = learningPipeline.AsDynamic;
// Create a binary classification trainer.
var binaryTrainer = new AveragedPerceptronTrainer(env, "Label", "Features");
// Append the OVA learner to the pipeline.
dynamicPipe = dynamicPipe.Append(new Ova(env, binaryTrainer));
// At this point, we have a choice. We could continue working with the dynamically-typed pipeline, and
// ultimately call dynamicPipe.Fit(data.AsDynamic) to get the model, or we could go back into the static world.
// Here's how we go back to the static pipeline:
var staticFinalPipe = dynamicPipe.AssertStatic(env,
// Declare the shape of the input. As you can see, it's identical to the shape of the reader:
// four float features and a string label.
c => (
SepalLength: c.R4.Scalar,
SepalWidth: c.R4.Scalar,
PetalLength: c.R4.Scalar,
PetalWidth: c.R4.Scalar,
Label: c.Text.Scalar),
// Declare the shape of the output (or a relevant subset of it).
// In our case, we care only about the predicted label column (a key type), and scores (vector of floats).
c => (
Score: c.R4.Vector,
// Predicted label is a key backed by uint, with text values (since original labels are text).
PredictedLabel: c.KeyU4.TextValues.Scalar))
// Convert the predicted label from key back to the original string value.
.Append(r => r.PredictedLabel.ToValue());
// Train the model in a statically typed way.
var model = staticFinalPipe.Fit(data);
// And here is how we could've stayed in the dynamic pipeline and train that way.
dynamicPipe = dynamicPipe.Append(new KeyToValueEstimator(env, "PredictedLabel"));
var dynamicModel = dynamicPipe.Fit(data.AsDynamic);
// Now 'dynamicModel', and 'model.AsDynamic' are equivalent.
var rs = model.Transform(data).GetColumn(x => x).ToArray();
}
