public double Learn(double[][] observations, int[] labels)
{
var gridsearch = GridSearch <double[], int> .Create(
ranges : new
{
Tolerance = GridSearch.Range(1e-10, 1.0, stepSize: 0.05)
},
learner : (p) => new LinearDualCoordinateDescent
{
Complexity = 1e+10,
Tolerance = p.Tolerance
},
fit : (teacher, x, y, w) => teacher.Learn(x, y, w),
loss : (actual, expected, m) => new ZeroOneLoss(expected).Loss(actual)
);
gridsearch.ParallelOptions.MaxDegreeOfParallelism = 2;
var result = gridsearch.Learn(observations, labels);
machine = result.BestModel;
bool[] output = machine.Decide(observations);
int[] zeroOneAnswers = output.ToZeroOne();
double ratio = 1 - (new AccuracyLoss(labels).Loss(zeroOneAnswers));
return(ratio);
}
public static void Train(string trainingFolder)
{
Console.WriteLine("Training SVM model with Cross-Validation...");
(double[][] inputs, int[] output) = ReadData(trainingFolder);
int crossValidateCount = Math.Min(maxCrossValidateCount, inputs.Count());
Accord.Math.Random.Generator.Seed = 0;
Console.WriteLine("Grid-Search...");
var gscv = GridSearch <double[], int> .CrossValidate(
ranges : new
{
Sigma = GridSearch.Range(fromInclusive: 0.00000001, toExclusive: 3),
},
learner : (p, ss) => new MulticlassSupportVectorLearning <Gaussian>
{
Kernel = new Gaussian(p.Sigma)
},
fit : (teacher, x, y, w) => teacher.Learn(x, y, w),
loss : (actual, expected, r) => new ZeroOneLoss(expected).Loss(actual),
folds : 10);
//gscv.ParallelOptions.MaxDegreeOfParallelism = 1;
var result = gscv.Learn(inputs.Take(crossValidateCount).ToArray(), output.Take(crossValidateCount).ToArray());
var crossValidation = result.BestModel;
double bestError = result.BestModelError;
double trainError = result.BestModel.Training.Mean;
double trainErrorVar = result.BestModel.Training.Variance;
double valError = result.BestModel.Validation.Mean;
double valErrorVar = result.BestModel.Validation.Variance;
double bestSigma = result.BestParameters.Sigma;
Console.WriteLine("Grid-Search Done.");
Console.WriteLine("Using Sigma=" + bestSigma);
// train model with best parameter
var bestTeacher = new MulticlassSupportVectorLearning <Gaussian>
{
Kernel = new Gaussian(bestSigma)
};
MulticlassSupportVectorMachine <Gaussian> svm = bestTeacher.Learn(
inputs.Take(traingRowsCount).ToArray(),
output.Take(traingRowsCount).ToArray());
// save model
svm.Save(Path.Combine(trainingFolder, "model"), SerializerCompression.GZip);
}
public void cross_validation_decision_tree()
{
#region doc_learn_tree_cv
// Ensure results are reproducible
Accord.Math.Random.Generator.Seed = 0;
// This is a sample code showing how to use Grid-Search in combination with
// Cross-Validation to assess the performance of Decision Trees with C4.5.
var parkinsons = new Parkinsons();
double[][] input = parkinsons.Features;
int[] output = parkinsons.ClassLabels;
// Create a new Grid-Search with Cross-Validation algorithm. Even though the
// generic, strongly-typed approach used accross the framework is most of the
// time easier to handle, combining those both methods in a single call can be
// difficult. For this reason. the framework offers a specialized method for
// combining those two algorirthms:
var gscv = GridSearch.CrossValidate(
// Here we can specify the range of the parameters to be included in the search
ranges: new
{
Join = GridSearch.Range(fromInclusive: 1, toExclusive: 20),
MaxHeight = GridSearch.Range(fromInclusive: 1, toExclusive: 20),
},
// Indicate how learning algorithms for the models should be created
learner: (p, ss) => new C45Learning
{
// Here, we can use the parameters we have specified above:
Join = p.Join,
MaxHeight = p.MaxHeight,
},
// Define how the model should be learned, if needed
fit: (teacher, x, y, w) => teacher.Learn(x, y, w),
// Define how the performance of the models should be measured
loss: (actual, expected, r) => new ZeroOneLoss(expected).Loss(actual),
folds: 3, // use k = 3 in k-fold cross validation
x: input, y: output // so the compiler can infer generic types
);
// If needed, control the parallelization degree
gscv.ParallelOptions.MaxDegreeOfParallelism = 1;
// Search for the best decision tree
var result = gscv.Learn(input, output);
// Get the best cross-validation result:
var crossValidation = result.BestModel;
// Get an estimate of its error:
double bestAverageError = result.BestModelError;
double trainError = result.BestModel.Training.Mean;
double trainErrorVar = result.BestModel.Training.Variance;
double valError = result.BestModel.Validation.Mean;
double valErrorVar = result.BestModel.Validation.Variance;
// Get the best values for the parameters:
int bestJoin = result.BestParameters.Join;
int bestHeight = result.BestParameters.MaxHeight;
// Use the best parameter values to create the final
// model using all the training and validation data:
var bestTeacher = new C45Learning
{
Join = bestJoin,
MaxHeight = bestHeight,
};
// Use the best parameters to create the final tree model:
DecisionTree finalTree = bestTeacher.Learn(input, output);
#endregion
int height = finalTree.GetHeight();
Assert.AreEqual(5, height);
Assert.AreEqual(22, result.BestModel.NumberOfInputs);
Assert.AreEqual(2, result.BestModel.NumberOfOutputs);
Assert.AreEqual(195, result.BestModel.NumberOfSamples);
Assert.AreEqual(65, result.BestModel.AverageNumberOfSamples);
Assert.AreEqual(bestAverageError, valError);
Assert.AreEqual(5, bestJoin, 1e-10);
Assert.AreEqual(0.1076923076923077, bestAverageError, 1e-8);
Assert.AreEqual(5, bestHeight, 1e-8);
}
public void learn_test_strongly_typed()
{
#region doc_learn_strongly_typed
// Ensure results are reproducible
Accord.Math.Random.Generator.Seed = 0;
// This is a sample code showing how to use Grid-Search in combination with
// Cross-Validation to assess the performance of Support Vector Machines.
// Consider the example binary data. We will be trying to learn a XOR
// problem and see how well does SVMs perform on this data.
double[][] inputs =
{
new double[] { -1, -1 }, new double[] { 1, -1 },
new double[] { -1, 1 }, new double[] { 1, 1 },
new double[] { -1, -1 }, new double[] { 1, -1 },
new double[] { -1, 1 }, new double[] { 1, 1 },
new double[] { -1, -1 }, new double[] { 1, -1 },
new double[] { -1, 1 }, new double[] { 1, 1 },
new double[] { -1, -1 }, new double[] { 1, -1 },
new double[] { -1, 1 }, new double[] { 1, 1 },
};
int[] xor = // result of xor for the sample input data
{
-1, 1,
1, -1,
-1, 1,
1, -1,
-1, 1,
1, -1,
-1, 1,
1, -1,
};
// Create a new Grid-Search with Cross-Validation algorithm. Even though the
// generic, strongly-typed approach used accross the framework is most of the
// time easier to handle, meta-algorithms such as grid-search can be a bit hard
// to setup. For this reason. the framework offers a specialized method for it:
var gridsearch = GridSearch <double[], int> .Create(
// Here we can specify the range of the parameters to be included in the search
ranges : new
{
Kernel = GridSearch.Values <IKernel>(new Linear(), new ChiSquare(), new Gaussian(), new Sigmoid()),
Complexity = GridSearch.Values(0.00000001, 5.20, 0.30, 0.50),
Tolerance = GridSearch.Range(1e-10, 1.0, stepSize: 0.05)
},
// Indicate how learning algorithms for the models should be created
learner : (p) => new SequentialMinimalOptimization <IKernel>
{
Complexity = p.Complexity,
Kernel = p.Kernel.Value,
Tolerance = p.Tolerance
},
// Define how the model should be learned, if needed
fit : (teacher, x, y, w) => teacher.Learn(x, y, w),
// Define how the performance of the models should be measured
loss : (actual, expected, m) => new ZeroOneLoss(expected).Loss(actual)
);
// If needed, control the degree of CPU parallelization
gridsearch.ParallelOptions.MaxDegreeOfParallelism = 1;
// Search for the best model parameters
var result = gridsearch.Learn(inputs, xor);
// Get the best SVM:
SupportVectorMachine <IKernel> svm = result.BestModel;
// Estimate its error:
double bestError = result.BestModelError;
// Get the best values for the parameters:
double bestC = result.BestParameters.Complexity;
double bestTolerance = result.BestParameters.Tolerance;
IKernel bestKernel = result.BestParameters.Kernel.Value;
#endregion
Assert.IsNotNull(svm);
Assert.AreEqual(1e-8, bestC, 1e-10);
Assert.AreEqual(0, bestError, 1e-8);
Assert.AreEqual(0, bestTolerance, 1e-8);
Assert.AreEqual(typeof(Gaussian), bestKernel.GetType());
}
public void learn_test_exception()
{
// https://github.com/accord-net/framework/issues/1052
Accord.Math.Random.Generator.Seed = 0;
double[][] inputs =
{
new double[] { -1, -1 }, new double[] { 1, -1 },
new double[] { -1, 1 }, new double[] { 1, 1 },
new double[] { -1, -1 }, new double[] { 1, -1 },
new double[] { -1, 1 }, new double[] { 1, 1 },
new double[] { -1, -1 }, new double[] { 1, -1 },
new double[] { -1, 1 }, new double[] { 1, 1 },
new double[] { -1, -1 }, new double[] { 1, -1 },
new double[] { -1, 1 }, new double[] { 1, 1 },
};
int[] xor = // result of xor for the sample input data
{
-1, 1,
1, -1,
-1, 1,
1, -1,
-1, 1,
1, -1,
-1, 1,
1, -1,
};
var gridsearch = GridSearch <double[], int> .Create(
ranges : new
{
Kernel = GridSearch.Values <IKernel>(new Linear()),
Complexity = GridSearch.Values(100000000000),
Tolerance = GridSearch.Range(1e-10, 1.0, stepSize: 0.05)
},
learner : (p) => new SequentialMinimalOptimization <IKernel>
{
Complexity = p.Complexity,
Kernel = p.Kernel.Value,
Tolerance = p.Tolerance
},
fit : (teacher, x, y, w) =>
{
try
{
return(teacher.Learn(x, y, w));
}
finally
{
throw new Exception("abacaxi");
}
},
loss : (actual, expected, m) => new ZeroOneLoss(expected).Loss(actual)
);
var result = gridsearch.Learn(inputs, xor);
SupportVectorMachine <IKernel> svm = result.BestModel;
double bestError = result.BestModelError;
double bestC = result.BestParameters.Complexity;
double bestTolerance = result.BestParameters.Tolerance;
IKernel bestKernel = result.BestParameters.Kernel.Value;
Assert.IsNull(svm);
Assert.AreEqual(20, result.Exceptions.Length);
foreach (Exception ex in result.Exceptions)
{
Assert.AreEqual("abacaxi", ex.Message);
}
Assert.AreEqual(100000000000, bestC, 1e-10);
Assert.AreEqual(Double.PositiveInfinity, bestError, 1e-8);
Assert.AreEqual(1E-10, bestTolerance, 1e-8);
Assert.AreEqual(typeof(Linear), bestKernel.GetType());
}
public void internals_test()
{
Accord.Math.Random.Generator.Seed = 0;
string[] inputs =
{
"input 1",
"input 2",
"input 3",
"input 4",
};
string[] outputs =
{
"output 1",
"output 2",
"output 3",
};
double[] weights =
{
1.0,
2.0,
3.0
};
var lossModels = new List <Mapper>();
var ranges = new
{
Parameter1 = GridSearch.Range("parameter 11", "parameter 12"),
Parameter2 = GridSearch.Range("parameter 21", "parameter 22", "parameter 23", "parameter 24"),
Parameter3 = GridSearch.Range("parameter 31")
};
var result = GridSearch.Create(
ranges: ranges,
learner: (p) => new MapperLearning
{
Parameter1 = p.Parameter1,
Parameter2 = p.Parameter2,
Parameter3 = p.Parameter3,
},
fit: (teacher, x, y, w) => teacher.Learn(x, y, w),
loss: (actual, expected, m) =>
{
if (m.Parameter1 == "parameter 12" && m.Parameter2 == "parameter 21" && m.Parameter3 == "parameter 31")
{
return(-42);
}
lock (lossModels)
{
lossModels.Add(m);
}
return(Math.Abs(int.Parse(m.Parameter1.Replace("parameter ", ""))
+ 100 * int.Parse(m.Parameter2.Replace("parameter ", ""))
+ 10000 * int.Parse(m.Parameter3.Replace("parameter ", ""))));
},
x: inputs,
y: outputs,
weights: weights
);
Mapper bestModel = result.BestModel;
Assert.AreEqual("parameter 12", bestModel.Parameter1);
Assert.AreEqual("parameter 21", bestModel.Parameter2);
Assert.AreEqual("parameter 31", bestModel.Parameter3);
Assert.AreEqual(inputs, bestModel.Inputs);
Assert.AreEqual(outputs, bestModel.Outputs);
Assert.AreEqual(weights, bestModel.Weights);
Assert.AreEqual(-42, result.BestModelError);
Assert.AreEqual(4, result.BestModelIndex);
var bestParameters = result.BestParameters;
Assert.AreNotSame(ranges, bestParameters);
Assert.AreEqual(1, bestParameters.Parameter1.Index);
Assert.AreEqual(0, bestParameters.Parameter2.Index);
Assert.AreEqual(0, bestParameters.Parameter3.Index);
Assert.AreEqual("parameter 12", bestParameters.Parameter1.Value);
Assert.AreEqual("parameter 21", bestParameters.Parameter2.Value);
Assert.AreEqual("parameter 31", bestParameters.Parameter3.Value);
Assert.AreEqual(8, result.Count);
Assert.AreEqual(result.Errors, new double[] {
312111, 312211, 312311, 312411,
-42, Double.PositiveInfinity, 312312, 312412
});
Exception[] exceptions = result.Exceptions;
for (int i = 0; i < exceptions.Length; i++)
{
if (i != 5)
{
Assert.IsNull(exceptions[i]);
}
else
{
Assert.AreEqual("Exception test", exceptions[i].Message);
}
}
Mapper[] models = result.Models;
Assert.AreEqual(8, models.Length);
Assert.AreEqual(6, lossModels.Count);
int a = ranges.Parameter1.Length;
int b = ranges.Parameter2.Length;
int c = ranges.Parameter3.Length;
Assert.AreEqual(2, a);
Assert.AreEqual(4, b);
Assert.AreEqual(1, c);
for (int i = 0; i < models.Length; i++)
{
if (i == 5)
{
Assert.IsNull(models[i]);
}
else
{
Assert.AreEqual(inputs, models[i].Inputs);
Assert.AreEqual(outputs, models[i].Outputs);
Assert.AreEqual(weights, models[i].Weights);
Assert.AreEqual(4, models[i].NumberOfInputs);
Assert.AreEqual(2, models[i].NumberOfOutputs);
Assert.AreEqual(ranges.Parameter1.Values[((i / c) / b) % a], models[i].Parameter1);
Assert.AreEqual(ranges.Parameter2.Values[(i / c) % b], models[i].Parameter2);
Assert.AreEqual(ranges.Parameter3.Values[i % c], models[i].Parameter3);
if (i != 4)
{
Assert.IsTrue(lossModels.Contains(models[i]));
}
}
}
Assert.AreEqual(4, result.NumberOfInputs);
Assert.AreEqual(2, result.NumberOfOutputs);
var parameters = result.Parameters;
for (int i = 0; i < parameters.Length; i++)
{
for (int j = 0; j < parameters.Length; j++)
{
if (i != j)
{
Assert.AreNotSame(parameters[i], parameters[j]);
Assert.AreNotEqual(parameters[i], parameters[j]);
Assert.AreNotEqual(parameters[i].Parameter1, parameters[j].Parameter1);
Assert.AreNotEqual(parameters[i].Parameter2, parameters[j].Parameter2);
Assert.AreNotEqual(parameters[i].Parameter3, parameters[j].Parameter3);
}
Assert.AreEqual(parameters[i].Parameter1.Values, parameters[j].Parameter1.Values);
Assert.AreEqual(parameters[i].Parameter2.Values, parameters[j].Parameter2.Values);
Assert.AreEqual(parameters[i].Parameter3.Values, parameters[j].Parameter3.Values);
}
}
}
public void cross_validation_test()
{
#region doc_learn_cv
// Ensure results are reproducible
Accord.Math.Random.Generator.Seed = 0;
// This is a sample code showing how to use Grid-Search in combination with
// Cross-Validation to assess the performance of Support Vector Machines.
// Consider the example binary data. We will be trying to learn a XOR
// problem and see how well does SVMs perform on this data.
double[][] inputs =
{
new double[] { -1, -1 }, new double[] { 1, -1 },
new double[] { -1, 1 }, new double[] { 1, 1 },
new double[] { -1, -1 }, new double[] { 1, -1 },
new double[] { -1, 1 }, new double[] { 1, 1 },
new double[] { -1, -1 }, new double[] { 1, -1 },
new double[] { -1, 1 }, new double[] { 1, 1 },
new double[] { -1, -1 }, new double[] { 1, -1 },
new double[] { -1, 1 }, new double[] { 1, 1 },
};
int[] xor = // result of xor for the sample input data
{
-1, 1,
1, -1,
-1, 1,
1, -1,
-1, 1,
1, -1,
-1, 1,
1, -1,
};
// Create a new Grid-Search with Cross-Validation algorithm. Even though the
// generic, strongly-typed approach used accross the framework is most of the
// time easier to handle, combining those both methods in a single call can be
// difficult. For this reason. the framework offers a specialized method for
// combining those two algorirthms:
var gscv = GridSearch <double[], int> .CrossValidate(
// Here we can specify the range of the parameters to be included in the search
ranges : new
{
Complexity = GridSearch.Range(new double[] { 0.00000001, 5.20, 0.30, 0.50 }),
Degree = GridSearch.Range(new int[] { 1, 10, 2, 3, 4, 5 }),
Constant = GridSearch.Range(new double[] { 0, 1, 2 }),
},
// Indicate how learning algorithms for the models should be created
learner : (p, ss) => new SequentialMinimalOptimization <Polynomial>
{
// Here, we can use the parameters we have specified above:
Complexity = p.Complexity,
Kernel = new Polynomial(p.Degree, p.Constant)
},
// Define how the model should be learned, if needed
fit : (teacher, x, y, w) => teacher.Learn(x, y, w),
// Define how the performance of the models should be measured
loss : (actual, expected, r) => new ZeroOneLoss(expected).Loss(actual),
folds : 3 // use k = 3 in k-fold cross validation
);
// If needed, control the parallelization degree
gscv.ParallelOptions.MaxDegreeOfParallelism = 1;
// Search for the best vector machine
var result = gscv.Learn(inputs, xor);
// Get the best cross-validation result:
var crossValidation = result.BestModel;
// Estimate its error:
double bestError = result.BestModelError;
double trainError = result.BestModel.Training.Mean;
double trainErrorVar = result.BestModel.Training.Variance;
double valError = result.BestModel.Validation.Mean;
double valErrorVar = result.BestModel.Validation.Variance;
// Get the best values for the parameters:
double bestC = result.BestParameters.Complexity;
double bestDegree = result.BestParameters.Degree;
double bestConstant = result.BestParameters.Constant;
#endregion
Assert.AreEqual(2, result.BestModel.NumberOfInputs);
Assert.AreEqual(1, result.BestModel.NumberOfOutputs);
Assert.AreEqual(16, result.BestModel.NumberOfSamples);
Assert.AreEqual(5.333333333333333, result.BestModel.AverageNumberOfSamples);
Assert.AreEqual(1e-8, bestC, 1e-10);
Assert.AreEqual(0, bestError, 1e-8);
Assert.AreEqual(10, bestDegree, 1e-8);
Assert.AreEqual(0, bestConstant, 1e-8);
}
