// Canvas events - obsolete
private void inputCanvas_MouseUp(object sender, MouseEventArgs e)
{
double[][] input = Sequence.Preprocess(Get3DSequence());
if (input.Length < 5)
{
panelUserLabeling.Visible = false;
panelClassification.Visible = false;
return;
}
if (hmm == null && hcrf == null)
{
panelUserLabeling.Visible = true;
panelClassification.Visible = false;
}
else
{
int index = (hcrf != null) ? hcrf.Compute(input) : hmm.Compute(input);
string label = database.Classes[index];
//int[] path = ;
lbHaveYouDrawn.Text = String.Format("Have you painted a {0}?", label);
panelClassification.Visible = true;
panelUserLabeling.Visible = false;
}
}
public void LearnTest1()
{
// Create a Continuous density Hidden Markov Model Sequence Classifier
// to detect a univariate sequence and the same sequence backwards.
double[][] sequences = new double[][]
{
new double[] { 0,1,2,3,4 }, // This is the first sequence with label = 0
new double[] { 4,3,2,1,0 }, // This is the second sequence with label = 1
};
// Labels for the sequences
int[] labels = { 0, 1 };
// Creates a sequence classifier containing 2 hidden Markov Models
// with 2 states and an underlying Normal distribution as density.
NormalDistribution density = new NormalDistribution();
var classifier = new HiddenMarkovClassifier<NormalDistribution>(2, new Ergodic(2), density);
// Configure the learning algorithms to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning<NormalDistribution>(classifier,
// Train each model until the log-likelihood changes less than 0.001
modelIndex => new BaumWelchLearning<NormalDistribution>(classifier.Models[modelIndex])
{
Tolerance = 0.0001,
Iterations = 0
}
);
// Train the sequence classifier using the algorithm
double logLikelihood = teacher.Run(sequences, labels);
// Calculate the probability that the given
// sequences originated from the model
double likelihood1, likelihood2;
// Try to classify the first sequence (output should be 0)
int c1 = classifier.Compute(sequences[0], out likelihood1);
// Try to classify the second sequence (output should be 1)
int c2 = classifier.Compute(sequences[1], out likelihood2);
Assert.AreEqual(0, c1);
Assert.AreEqual(1, c2);
Assert.AreEqual(-13.271981026832929, logLikelihood, 1e-10);
Assert.AreEqual(0.99999791320102149, likelihood1, 1e-10);
Assert.AreEqual(0.99999791320102149, likelihood2, 1e-10);
Assert.IsFalse(double.IsNaN(logLikelihood));
Assert.IsFalse(double.IsNaN(likelihood1));
Assert.IsFalse(double.IsNaN(likelihood2));
}
public void LearnTest1()
{
// Create a Continuous density Hidden Markov Model Sequence Classifier
// to detect a univariate sequence and the same sequence backwards.
double[][] sequences = new double[][]
{
new double[] { 0, 1, 2, 3, 4 }, // This is the first sequence with label = 0
new double[] { 4, 3, 2, 1, 0 }, // This is the second sequence with label = 1
};
// Labels for the sequences
int[] labels = { 0, 1 };
// Creates a sequence classifier containing 2 hidden Markov Models
// with 2 states and an underlying Normal distribution as density.
NormalDistribution density = new NormalDistribution();
var classifier = new HiddenMarkovClassifier <NormalDistribution>(2, new Ergodic(2), density);
// Configure the learning algorithms to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning <NormalDistribution>(classifier,
// Train each model until the log-likelihood changes less than 0.001
modelIndex => new BaumWelchLearning <NormalDistribution>(classifier.Models[modelIndex])
{
Tolerance = 0.0001,
Iterations = 0
}
);
// Train the sequence classifier using the algorithm
double logLikelihood = teacher.Run(sequences, labels);
// Calculate the probability that the given
// sequences originated from the model
double likelihood1, likelihood2;
// Try to classify the first sequence (output should be 0)
int c1 = classifier.Compute(sequences[0], out likelihood1);
// Try to classify the second sequence (output should be 1)
int c2 = classifier.Compute(sequences[1], out likelihood2);
Assert.AreEqual(0, c1);
Assert.AreEqual(1, c2);
Assert.AreEqual(-13.271981026832929, logLikelihood, 1e-10);
Assert.AreEqual(0.99999791320102149, likelihood1, 1e-10);
Assert.AreEqual(0.99999791320102149, likelihood2, 1e-10);
Assert.IsFalse(double.IsNaN(logLikelihood));
Assert.IsFalse(double.IsNaN(likelihood1));
Assert.IsFalse(double.IsNaN(likelihood2));
}
public static void LearnAndPredictContinuous()
{
// Create a Continuous density Hidden Markov Model Sequence Classifier
// to detect a univariate sequence and the same sequence backwards.
double[][] sequences = new double[][]
{
new double[] { 0, 1, 2, 3, 4 }, // This is the first sequence with label = 0
new double[] { 4, 3, 2, 1, 0 }, // This is the second sequence with label = 1
};
// Labels for the sequences
int[] labels = { 0, 1 };
// Creates a new Continuous-density Hidden Markov Model Sequence Classifier
// containing 2 hidden Markov Models with 2 states and an underlying Normal
// distribution as the continuous probability density.
Gaussian density = new Gaussian();
var classifier = new HiddenMarkovClassifier(2, new Ergodic(2), density);
// Create a new learning algorithm to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning(classifier,
// Train each model until the log-likelihood changes less than 0.001
modelIndex => new BaumWelchLearning(classifier.Models[modelIndex])
{
Tolerance = 0.0001,
Iterations = 0
}
);
// Train the sequence classifier using the algorithm
teacher.Run(sequences, labels);
// Calculate the probability that the given
// sequences originated from the model
double likelihood;
// Try to classify the first sequence (output should be 0)
int c1 = classifier.Compute(sequences[0], out likelihood);
Console.WriteLine("c1: {0}", c1);
// Try to classify the second sequence (output should be 1)
int c2 = classifier.Compute(sequences[1], out likelihood);
Console.WriteLine("c2: {0}", c2);
}
public int Classify(double[][][] trainDataSet, int[] trainLabels, double[][] testData, String[] classes)
{
int states = 5;
int dimensionsOfFeatures = 12;
int numberOfClasses = classes.Length;
int iterations = 0;
double tolerance = 0.01;
HiddenMarkovClassifier <MultivariateNormalDistribution> hmm = new HiddenMarkovClassifier <MultivariateNormalDistribution>
(numberOfClasses, new Forward(states), new MultivariateNormalDistribution(dimensionsOfFeatures), classes);
// Create the learning algorithm for the ensemble classifier
var teacher = new HiddenMarkovClassifierLearning <MultivariateNormalDistribution>(hmm,
// Train each model using the selected convergence criteria
i => new BaumWelchLearning <MultivariateNormalDistribution>(hmm.Models[i])
{
Tolerance = tolerance,
Iterations = iterations,
FittingOptions = new NormalOptions()
{
Regularization = 1e-5
}
}
);
teacher.Empirical = true;
teacher.Rejection = false;
// Run the learning algorithm
double error = teacher.Run(trainDataSet, trainLabels);
int predictedResult = hmm.Compute(testData);
return(predictedResult);
}
public int Recognize(ISoundSignalReader signal, HiddenMarkovClassifier hmm, out string name,
SignalVisitor voiceVisitor = null)
{
var featureUtility = new FeatureUtility(_engineParameters);
signal.Reset();
var features = featureUtility.ExtractFeatures(signal, voiceVisitor).First();
var observations = _codeBook.Quantize(features.Select(item => new Point(item)).ToArray());
double[] responsabilities;
var ret = hmm.Compute(observations, out responsabilities);
var models = hmm.Models;
var likelyHoodValue = Double.MinValue;
name = string.Empty;
foreach (var model in models)
{
var val = model.Evaluate(observations);
if (val > likelyHoodValue)
{
likelyHoodValue = val;
name = model.Tag.ToString();
}
}
return ret;
}
private void classify(Canvas canvas, ZedGraphControl graph,
System.Windows.Forms.Label output)
{
int[] sequence = GetFeatures(canvas.GetSequence());
if (classifier == null || sequence.Length <= 10)
{
output.Text = "-";
return;
}
double[] responses;
int index = classifier.Compute(sequence, out responses);
output.Text = (index == -1) ?
"-" : classifier.Models[index].Tag as string;
// Scale the responses to a [0,1] interval
double max = responses.Max();
double min = responses.Min();
for (int i = 0; i < responses.Length; i++)
{
responses[i] = Accord.Math.Tools.Scale(min, max, 0, 1, responses[i]);
}
// Create the bar graph to show the relative responses
CreateBarGraph(graph, responses);
}
private void button1_Click(object sender, EventArgs e)
{
Point[] P = new Point[30000];
for (int i = 0; i < points.Count; i++)
{
P[i] = points[i];
richTextBox1.AppendText(P[i].ToString());
}
double[][] input = Sequence.Preprocess(P);
//if (input.Length < 5)
//{
// panelUserLabeling.Visible = false;
// panelClassification.Visible = false;
// return;
//}
if (hmm == null && hcrf == null)
{
panelUserLabeling.Visible = true;
panelClassification.Visible = false;
}
else
{
int index = (hcrf != null) ?
hcrf.Compute(input) : hmm.Compute(input);
string label = database.Classes[index];
lbHaveYouDrawn.Text = String.Format("Have you drawn a {0}?", label);
panelClassification.Visible = true;
panelUserLabeling.Visible = false;
}
Application.Idle -= addCoord;
}
public void LearnTest()
{
// Declare some testing data
int[][] inputs = new int[][]
{
new int[] { 0,1,1,0 }, // Class 0
new int[] { 0,0,1,0 }, // Class 0
new int[] { 0,1,1,1,0 }, // Class 0
new int[] { 0,1,0 }, // Class 0
new int[] { 1,0,0,1 }, // Class 1
new int[] { 1,1,0,1 }, // Class 1
new int[] { 1,0,0,0,1 }, // Class 1
new int[] { 1,0,1 }, // Class 1
};
int[] outputs = new int[]
{
0,0,0,0, // First four sequences are of class 0
1,1,1,1, // Last four sequences are of class 1
};
// We are trying to predict two different classes
int classes = 2;
// Each sequence may have up to two symbols (0 or 1)
int symbols = 2;
// Nested models will have two states each
int[] states = new int[] { 2, 2 };
// Creates a new Hidden Markov Model Classifier with the given parameters
HiddenMarkovClassifier classifier = new HiddenMarkovClassifier(classes, states, symbols);
// Create a new learning algorithm to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning(classifier,
// Train each model until the log-likelihood changes less than 0.001
modelIndex => new BaumWelchLearning(classifier.Models[modelIndex])
{
Tolerance = 0.001,
Iterations = 0
}
);
// Train the sequence classifier using the algorithm
double likelihood = teacher.Run(inputs, outputs);
// Will assert the models have learned the sequences correctly.
for (int i = 0; i < inputs.Length; i++)
{
int expected = outputs[i];
int actual = classifier.Compute(inputs[i], out likelihood);
Assert.AreEqual(expected, actual);
}
}
public void Run()
{
/*Initialize the model
* Read more tut on Code project for better understanding
* http://www.codeproject.com/Articles/541428/Sequence-Classifiers-in-Csharp-Part-I-Hidden-Marko?msg=5219822#xx5219822xx
* states is parameters for running forward algo
* intteration is parameters for iterations
* tolerance is parameters for threshold
* */
int states = 3;
int iterations = 100;
double tolerance = 0.01;
bool rejection = false;
string[] classes = ActivityIndex.Keys.ToArray();
ITopology foward = new Forward(states: 3);
hmm = new HiddenMarkovClassifier(classes: 12, topology: foward, symbols: 5);
// Create the learning algorithm for the ensemble classifier
var teacher = new HiddenMarkovClassifierLearning(hmm,
// Train each model using the selected convergence criteria
i => new BaumWelchLearning(hmm.Models[i])
{
Tolerance = tolerance,
Iterations = iterations,
}
);
teacher.Empirical = true;
teacher.Rejection = rejection;
// Run the learning algorithm
double error = teacher.Run(input, output);
Console.WriteLine("Error: {0}", error);
//Run the test and compare the real value
using (StreamWriter writer = new StreamWriter("compare.txt"))
{
for (int i = 0; i < outpuTest.Length; ++i)
{
int val = hmm.Compute(inpuTest[i]);
if (val != outpuTest[i])
{
string labelTestRetrieve = ActivityIndex.FirstOrDefault(x => x.Value == val).Key;
string labelTestActity = ActivityIndex.FirstOrDefault(x => x.Value == outpuTest[i]).Key;
writer.WriteLine(outputTestLabelFolder[i] + " - " + "false, label test retrieve: " + labelTestRetrieve + " label activity: " + labelTestActity);
}
else
{
string labelTestActity = ActivityIndex.FirstOrDefault(x => x.Value == outpuTest[i]).Key;
writer.WriteLine(outputTestLabelFolder[i] + " - " + "true, label activity: " + labelTestActity);
}
}
}
}
/// <summary>
/// learn Hmm model for samples in given database using Baum Welch unsupervised learning algorithm
/// then used the learned model to classify the training samples
/// </summary>
/// <param name="database"></param>
/// <returns></returns>
static HiddenMarkovClassifier <MultivariateNormalDistribution> learnHMM(Database database)
{
BindingList <Sequence> samples = database.Samples;
BindingList <String> classes = database.Classes;
double[][][] inputs = new double[samples.Count][][];
int[] outputs = new int[samples.Count];
for (int i = 0; i < inputs.Length; i++)
{
inputs[i] = samples[i].Input;
outputs[i] = samples[i].Output;
}
int states = 5;
int iterations = 0;
double tolerance = 0.1;
bool rejection = true;
HiddenMarkovClassifier <MultivariateNormalDistribution> hmm = new HiddenMarkovClassifier <MultivariateNormalDistribution>(classes.Count,
new Forward(states), new MultivariateNormalDistribution(2), classes.ToArray());
// Create the learning algorithm for the ensemble classifier
var teacher = new HiddenMarkovClassifierLearning <MultivariateNormalDistribution>(hmm,
// Train each model using the selected convergence criteria
i => new BaumWelchLearning <MultivariateNormalDistribution>(hmm.Models[i])
{
Tolerance = tolerance,
Iterations = iterations,
FittingOptions = new NormalOptions()
{
Regularization = 1e-5
}
}
);
teacher.Empirical = true;
teacher.Rejection = rejection;
// Run the learning algorithm
double error = teacher.Run(inputs, outputs);
// Classify all training instances
foreach (var sample in database.Samples)
{
sample.RecognizedAs = hmm.Compute(sample.Input);
}
return(hmm);
}
private static void check4(double[][][] words, HiddenMarkovClassifier <Independent> model, MarkovMultivariateFunction target, HiddenConditionalRandomField <double[]> hcrf)
{
double actual;
double expected;
foreach (var x in words)
{
for (int c = 0; c < model.Classes; c++)
{
for (int i = 0; i < model[c].States; i++)
{
// Check initial state transitions
double xa = model.Priors[c];
double xb = Math.Exp(model[c].Probabilities[i]);
double xc = model[c].Emissions[i].ProbabilityDensityFunction(x[0]);
expected = xa * xb * xc;
actual = Math.Exp(target.Factors[c].Compute(-1, i, x, 0, c));
Assert.IsTrue(expected.IsRelativelyEqual(actual, 1e-10));
Assert.IsFalse(double.IsNaN(actual));
}
for (int t = 1; t < x.Length; t++)
{
// Check normal state transitions
for (int i = 0; i < model[c].States; i++)
{
for (int j = 0; j < model[c].States; j++)
{
double xb = Math.Exp(model[c].Transitions[i, j]);
double xc = model[c].Emissions[j].ProbabilityDensityFunction(x[t]);
expected = xb * xc;
actual = Math.Exp(target.Factors[c].Compute(i, j, x, t, c));
Assert.IsTrue(expected.IsRelativelyEqual(actual, 1e-10));
Assert.IsFalse(double.IsNaN(actual));
}
}
}
actual = Math.Exp(model.LogLikelihood(x, c));
expected = Math.Exp(hcrf.LogLikelihood(x, c));
Assert.AreEqual(expected, actual, 1e-10);
Assert.IsFalse(double.IsNaN(actual));
actual = model.Compute(x);
expected = hcrf.Compute(x);
Assert.AreEqual(expected, actual);
Assert.IsFalse(double.IsNaN(actual));
}
}
}
public void ComputeTest()
{
HiddenMarkovClassifier hmm = DiscreteHiddenMarkovClassifierPotentialFunctionTest.CreateModel1();
// Declare some testing data
int[][] inputs = new int[][]
{
new int[] { 0, 1, 1, 0 }, // Class 0
new int[] { 0, 0, 1, 0 }, // Class 0
new int[] { 0, 1, 1, 1, 0 }, // Class 0
new int[] { 0, 1, 0 }, // Class 0
new int[] { 1, 0, 0, 1 }, // Class 1
new int[] { 1, 1, 0, 1 }, // Class 1
new int[] { 1, 0, 0, 0, 1 }, // Class 1
new int[] { 1, 0, 1 }, // Class 1
};
int[] outputs = new int[]
{
0, 0, 0, 0, // First four sequences are of class 0
1, 1, 1, 1, // Last four sequences are of class 1
};
var function = new MarkovDiscreteFunction(hmm);
var target = new HiddenConditionalRandomField <int>(function);
for (int i = 0; i < inputs.Length; i++)
{
int expected = hmm.Compute(inputs[i]);
int actual = target.Compute(inputs[i]);
double h0 = hmm.LogLikelihood(inputs[i], 0);
double h1 = hmm.LogLikelihood(inputs[i], 1);
double c0 = target.LogLikelihood(inputs[i], 0);
double c1 = target.LogLikelihood(inputs[i], 1);
Assert.AreEqual(expected, actual);
Assert.AreEqual(h0, c0, 1e-10);
Assert.AreEqual(h1, c1, 1e-10);
Assert.IsFalse(double.IsNaN(c0));
Assert.IsFalse(double.IsNaN(c1));
}
}
public double ComputeError(int[][] inputs, int[] outputs)
{
int errors = 0;
Parallel.For(0, inputs.Length, i =>
{
int expectedOutput = outputs[i];
int actualOutput = mClassifier.Compute(inputs[i]);
if (expectedOutput != actualOutput)
{
Interlocked.Increment(ref errors);
}
});
return(errors / (double)inputs.Length);
}
/// <summary>
/// Tests the ensemble
/// </summary>
private void btnTest_Click(object sender, EventArgs e)
{
int rows = dgvTesting.Rows.Count - 1;
// Gets the input sequences
int[][] sequences = new int[rows][];
for (int i = 0; i < rows; i++)
{
DataGridViewRow row = dgvTesting.Rows[i];
string sequence = row.Cells["colTestSequence"].Value as string;
String trueLabel = row.Cells["colTestTrueClass"].Value as string;
double likelihood;
int label = hmmc.Compute(decode(sequence), out likelihood);
string assignedLabel = hmmc.Models[label].Tag as string;
row.Cells["colTestAssignedClass"].Value = assignedLabel;
row.Cells["colTestLikelihood"].Value = likelihood;
row.Cells["colTestMatch"].Value = trueLabel == assignedLabel;
}
}
/// <summary>
/// Tests the ensemble
/// </summary>
private void btnTest_Click(object sender, EventArgs e)
{
int rows = dgvTesting.Rows.Count - 1;
int[] expected = new int[rows];
int[] actual = new int[rows];
// Gets the input sequences
int[][] sequences = new int[rows][];
// For each row in the testing data
for (int i = 0; i < rows; i++)
{
// Get the row at the current index
DataGridViewRow row = dgvTesting.Rows[i];
// Get the training sequence to feed the model
int[] sequence = decode(row.Cells["colTestSequence"].Value as string);
// Get the label associated with this sequence
string label = row.Cells["colTestTrueClass"].Value as string;
expected[i] = hmmc.Models.Find(x => x.Tag as string == label)[0];
// Compute the model output for this sequence and
// get its associated likelihood.
double likelihood;
actual[i] = hmmc.Compute(sequence, out likelihood);
row.Cells["colTestAssignedClass"].Value = hmmc.Models[actual[i]].Tag as string;
row.Cells["colTestLikelihood"].Value = likelihood;
row.Cells["colTestMatch"].Value = actual[i] == expected[i];
}
// Use confusion matrix to compute some performance metrics
dgvPerformance.DataSource = new[] { new GeneralConfusionMatrix(hmmc.Classes, actual, expected) };
}
private void button1_Click(object sender, EventArgs e)
{
var classes = 4;
var states = new[] { 1, 2, 2, 3 };
var cat = new[] { "ខ្ញុំ", "ទៅ", "ខ្លួន", "ក" };
//var cat = new[] { "A", "B" };
_hmmc = new HiddenMarkovClassifier(classes, states, 4, cat);
// Train the ensemble
var sequences = new[]
{
new[] { 1, 1, 1 },
new[] { 0, 2 },
new[] { 0, 1, 2 },
new[] { 1, 2 }
};
var labels = new[] { 0, 1, 2, 3 };
var teacher = new HiddenMarkovClassifierLearning(_hmmc, i =>
new BaumWelchLearning(_hmmc.Models[i])
{
Iterations = 0,
Tolerance = 0.0001
}
);
teacher.Run(sequences, labels);
var m = _hmmc.Models;
var test = new[] { 1, 2 };
double likelihood;
var label = _hmmc.Compute(test, out likelihood);
MessageBox.Show(_hmmc.Models[label].Tag.ToString() + " P =" + likelihood);
}
private void button1_Click(object sender, EventArgs e)
{
var classes = 4;
var states = new[]{1,2,2,3};
var cat = new[] {"ខ្ញុំ", "ទៅ", "ខ្លួន", "ក"};
//var cat = new[] { "A", "B" };
_hmmc = new HiddenMarkovClassifier(classes, states, 4, cat);
// Train the ensemble
var sequences = new[]
{
new[] {1, 1, 1},
new[] {0, 2},
new[] {0, 1, 2},
new[] {1, 2}
};
var labels = new[] {0, 1, 2, 3};
var teacher = new HiddenMarkovClassifierLearning(_hmmc, i =>
new BaumWelchLearning(_hmmc.Models[i])
{
Iterations = 0,
Tolerance = 0.0001
}
);
teacher.Run(sequences, labels);
var m = _hmmc.Models;
var test = new[]{1,2};
double likelihood;
var label = _hmmc.Compute(test, out likelihood);
MessageBox.Show(_hmmc.Models[label].Tag.ToString()+ " P =" + likelihood);
}
public void SimpleGestureRecognitionTest()
{
// Let's say we would like to do a very simple mechanism for
// gesture recognition. In this example, we will be trying to
// create a classifier that can distinguish between the words
// "hello", "car", and "wardrobe".
// Let's say we decided to acquire some data, and we asked some
// people to perform those words in front of a Kinect camera, and,
// using Microsoft's SDK, we were able to captured the x and y
// coordinates of each hand while the word was being performed.
// Let's say we decided to represent our frames as:
//
// double[] frame = { leftHandX, leftHandY, rightHandX, rightHandY };
//
// Since we captured words, this means we captured sequences of
// frames as we described above. Let's write some of those as
// rough examples to explain how gesture recognition can be done:
double[][] hello =
{
new double[] { 1.0, 0.1, 0.0, 0.0 }, // let's say the word
new double[] { 0.0, 1.0, 0.1, 0.1 }, // hello took 6 frames
new double[] { 0.0, 1.0, 0.1, 0.1 }, // to be recorded.
new double[] { 0.0, 0.0, 1.0, 0.0 },
new double[] { 0.0, 0.0, 1.0, 0.0 },
new double[] { 0.0, 0.0, 0.1, 1.1 },
};
double[][] car =
{
new double[] { 0.0, 0.0, 0.0, 1.0 }, // the car word
new double[] { 0.1, 0.0, 1.0, 0.1 }, // took only 4.
new double[] { 0.0, 0.0, 0.1, 0.0 },
new double[] { 1.0, 0.0, 0.0, 0.0 },
};
double[][] wardrobe =
{
new double[] { 0.0, 0.0, 1.0, 0.0 }, // same for the
new double[] { 0.1, 0.0, 1.0, 0.1 }, // wardrobe word.
new double[] { 0.0, 0.1, 1.0, 0.0 },
new double[] { 0.1, 0.0, 1.0, 0.1 },
};
// Here, please note that a real-world example would involve *lots*
// of samples for each word. Here, we are considering just one from
// each class which is clearly sub-optimal and should _never_ be done
// on practice. For example purposes, however, please disregard this.
// Those are the words we have in our vocabulary:
//
double[][][] words = { hello, car, wardrobe };
// Now, let's associate integer labels with them. This is needed
// for the case where there are multiple samples for each word.
//
int[] labels = { 0, 1, 2 };
// We will create our classifiers assuming an independent
// Gaussian distribution for each component in our feature
// vectors (like assuming a Naive Bayes assumption).
var initial = new Independent<NormalDistribution>
(
new NormalDistribution(0, 1),
new NormalDistribution(0, 1),
new NormalDistribution(0, 1),
new NormalDistribution(0, 1)
);
// Now, we can proceed and create our classifier.
//
int numberOfWords = 3; // we are trying to distinguish between 3 words
int numberOfStates = 5; // this value can be found by trial-and-error
var hmm = new HiddenMarkovClassifier<Independent<NormalDistribution>>
(
classes: numberOfWords,
topology: new Forward(numberOfStates), // word classifiers should use a forward topology
initial: initial
);
// Create a new learning algorithm to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning<Independent<NormalDistribution>>(hmm,
// Train each model until the log-likelihood changes less than 0.001
modelIndex => new BaumWelchLearning<Independent<NormalDistribution>>(hmm.Models[modelIndex])
{
Tolerance = 0.001,
Iterations = 100,
// This is necessary so the code doesn't blow up when it realize
// there is only one sample per word class. But this could also be
// needed in normal situations as well.
//
FittingOptions = new IndependentOptions()
{
InnerOption = new NormalOptions() { Regularization = 1e-5 }
}
}
);
// Finally, we can run the learning algorithm!
double logLikelihood = teacher.Run(words, labels);
// At this point, the classifier should be successfully
// able to distinguish between our three word classes:
//
int tc1 = hmm.Compute(hello);
int tc2 = hmm.Compute(car);
int tc3 = hmm.Compute(wardrobe);
Assert.AreEqual(0, tc1);
Assert.AreEqual(1, tc2);
Assert.AreEqual(2, tc3);
// Now, we can use the Markov classifier to initialize a HCRF
var function = new MarkovMultivariateFunction(hmm);
var hcrf = new HiddenConditionalRandomField<double[]>(function);
// We can check that both are equivalent, although they have
// formulations that can be learned with different methods
//
for (int i = 0; i < words.Length; i++)
{
// Should be the same
int expected = hmm.Compute(words[i]);
int actual = hcrf.Compute(words[i]);
// Should be the same
double h0 = hmm.LogLikelihood(words[i], 0);
double c0 = hcrf.LogLikelihood(words[i], 0);
double h1 = hmm.LogLikelihood(words[i], 1);
double c1 = hcrf.LogLikelihood(words[i], 1);
double h2 = hmm.LogLikelihood(words[i], 2);
double c2 = hcrf.LogLikelihood(words[i], 2);
Assert.AreEqual(expected, actual);
Assert.AreEqual(h0, c0, 1e-10);
Assert.IsTrue(h1.IsRelativelyEqual(c1, 1e-10));
Assert.IsTrue(h2.IsRelativelyEqual(c2, 1e-10));
Assert.IsFalse(double.IsNaN(c0));
Assert.IsFalse(double.IsNaN(c1));
Assert.IsFalse(double.IsNaN(c2));
}
// Now we can learn the HCRF using one of the best learning
// algorithms available, Resilient Backpropagation learning:
// Create a learning algorithm
var rprop = new HiddenResilientGradientLearning<double[]>(hcrf)
{
Iterations = 50,
Tolerance = 1e-5
};
// Run the algorithm and learn the models
double error = rprop.Run(words, labels);
// At this point, the HCRF should be successfully
// able to distinguish between our three word classes:
//
int hc1 = hcrf.Compute(hello);
int hc2 = hcrf.Compute(car);
int hc3 = hcrf.Compute(wardrobe);
Assert.AreEqual(0, hc1);
Assert.AreEqual(1, hc2);
Assert.AreEqual(2, hc3);
}
private void btnLearnHMM_Click(object sender, EventArgs e)
{
if (gridSamples.Rows.Count == 0)
{
MessageBox.Show("Please load or insert some data first.");
return;
}
BindingList<Sequence> samples = database.Samples;
BindingList<String> classes = database.Classes;
double[][][] inputs = new double[samples.Count][][];
int[] outputs = new int[samples.Count];
for (int i = 0; i < inputs.Length; i++)
{
inputs[i] = samples[i].Input;
outputs[i] = samples[i].Output;
}
int states = 5;
int iterations = 0;
double tolerance = 0.01;
bool rejection = false;
hmm = new HiddenMarkovClassifier<MultivariateNormalDistribution>(classes.Count,
new Forward(states), new MultivariateNormalDistribution(2), classes.ToArray());
// Create the learning algorithm for the ensemble classifier
var teacher = new HiddenMarkovClassifierLearning<MultivariateNormalDistribution>(hmm,
// Train each model using the selected convergence criteria
i => new BaumWelchLearning<MultivariateNormalDistribution>(hmm.Models[i])
{
Tolerance = tolerance,
Iterations = iterations,
FittingOptions = new NormalOptions()
{
Regularization = 1e-5
}
}
);
teacher.Empirical = true;
teacher.Rejection = rejection;
// Run the learning algorithm
double error = teacher.Run(inputs, outputs);
// Classify all training instances
foreach (var sample in database.Samples)
{
sample.RecognizedAs = hmm.Compute(sample.Input);
}
foreach (DataGridViewRow row in gridSamples.Rows)
{
var sample = row.DataBoundItem as Sequence;
row.DefaultCellStyle.BackColor = (sample.RecognizedAs == sample.Output) ?
Color.LightGreen : Color.White;
}
btnLearnHCRF.Enabled = true;
}
public void LearnTest2()
{
// Declare some testing data
int[][] inputs = new int[][]
{
new int[] { 0,0,1,2 }, // Class 0
new int[] { 0,1,1,2 }, // Class 0
new int[] { 0,0,0,1,2 }, // Class 0
new int[] { 0,1,2,2,2 }, // Class 0
new int[] { 2,2,1,0 }, // Class 1
new int[] { 2,2,2,1,0 }, // Class 1
new int[] { 2,2,2,1,0 }, // Class 1
new int[] { 2,2,2,2,1 }, // Class 1
};
int[] outputs = new int[]
{
0,0,0,0, // First four sequences are of class 0
1,1,1,1, // Last four sequences are of class 1
};
// We are trying to predict two different classes
int classes = 2;
// Each sequence may have up to 3 symbols (0,1,2)
int symbols = 3;
// Nested models will have 3 states each
int[] states = new int[] { 3, 3 };
// Creates a new Hidden Markov Model Classifier with the given parameters
HiddenMarkovClassifier classifier = new HiddenMarkovClassifier(classes, states, symbols);
// Create a new learning algorithm to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning(classifier,
// Train each model until the log-likelihood changes less than 0.001
modelIndex => new BaumWelchLearning(classifier.Models[modelIndex])
{
Tolerance = 0.001,
Iterations = 0
}
);
// Enable support for sequence rejection
teacher.Rejection = true;
// Train the sequence classifier using the algorithm
double likelihood = teacher.Run(inputs, outputs);
// Will assert the models have learned the sequences correctly.
for (int i = 0; i < inputs.Length; i++)
{
int expected = outputs[i];
int actual = classifier.Compute(inputs[i], out likelihood);
Assert.AreEqual(expected, actual);
}
HiddenMarkovModel threshold = classifier.Threshold;
Assert.AreEqual(6, threshold.States);
Assert.AreEqual(classifier.Models[0].Transitions[0, 0], threshold.Transitions[0, 0], 1e-10);
Assert.AreEqual(classifier.Models[0].Transitions[1, 1], threshold.Transitions[1, 1], 1e-10);
Assert.AreEqual(classifier.Models[0].Transitions[2, 2], threshold.Transitions[2, 2], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[0, 0], threshold.Transitions[3, 3], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[1, 1], threshold.Transitions[4, 4], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[2, 2], threshold.Transitions[5, 5], 1e-10);
Assert.IsFalse(Matrix.HasNaN(threshold.Transitions));
int[] r0 = new int[] { 1, 1, 0, 0, 2 };
double logRejection;
int c = classifier.Compute(r0, out logRejection);
Assert.AreEqual(-1, c);
Assert.AreEqual(0.99569011079012049, logRejection);
Assert.IsFalse(double.IsNaN(logRejection));
logRejection = threshold.Evaluate(r0);
Assert.AreEqual(-6.7949285513628528, logRejection, 1e-10);
Assert.IsFalse(double.IsNaN(logRejection));
threshold.Decode(r0, out logRejection);
Assert.AreEqual(-8.902077561009957, logRejection, 1e-10);
Assert.IsFalse(double.IsNaN(logRejection));
}
static void runDiscreteDensityHiddenMarkovClassifierLearningExample()
{
// Observation sequences should only contain symbols that are greater than or equal to 0, and lesser than the number of symbols.
int[][] observationSequences =
{
// First class of sequences: starts and ends with zeros, ones in the middle.
new[] { 0, 1, 1, 1, 0 },
new[] { 0, 0, 1, 1, 0, 0 },
new[] { 0, 1, 1, 1, 1, 0 },
// Second class of sequences: starts with twos and switches to ones until the end.
new[] { 2, 2, 2, 2, 1, 1, 1, 1, 1 },
new[] { 2, 2, 1, 2, 1, 1, 1, 1, 1 },
new[] { 2, 2, 2, 2, 2, 1, 1, 1, 1 },
// Third class of sequences: can start with any symbols, but ends with three.
new[] { 0, 0, 1, 1, 3, 3, 3, 3 },
new[] { 0, 0, 0, 3, 3, 3, 3 },
new[] { 1, 0, 1, 2, 2, 2, 3, 3 },
new[] { 1, 1, 2, 3, 3, 3, 3 },
new[] { 0, 0, 1, 1, 3, 3, 3, 3 },
new[] { 2, 2, 0, 3, 3, 3, 3 },
new[] { 1, 0, 1, 2, 3, 3, 3, 3 },
new[] { 1, 1, 2, 3, 3, 3, 3 },
};
// Consider their respective class labels.
// Class labels have to be zero-based and successive integers.
int[] classLabels =
{
0, 0, 0, // Sequences 1-3 are from class 0.
1, 1, 1, // Sequences 4-6 are from class 1.
2, 2, 2, 2, 2, 2, 2, 2 // Sequences 7-14 are from class 2.
};
// Use a single topology for all inner models.
ITopology forward = new Forward(states: 3);
// Create a hidden Markov classifier with the given topology.
HiddenMarkovClassifier hmc = new HiddenMarkovClassifier(classes: 3, topology: forward, symbols: 4);
// Create a algorithms to teach each of the inner models.
var trainer = new HiddenMarkovClassifierLearning(
hmc,
// Specify individual training options for each inner model.
modelIndex => new BaumWelchLearning(hmc.Models[modelIndex])
{
Tolerance = 0.001, // iterate until log-likelihood changes less than 0.001.
Iterations = 0 // don't place an upper limit on the number of iterations.
}
);
// Call its Run method to start learning.
double averageLogLikelihood = trainer.Run(observationSequences, classLabels);
Console.WriteLine("average log-likelihood for the observations = {0}", averageLogLikelihood);
// Check the output classificaton label for some sequences.
int y1 = hmc.Compute(new[] { 0, 1, 1, 1, 0 }); // output is y1 = 0.
Console.WriteLine("output class = {0}", y1);
int y2 = hmc.Compute(new[] { 0, 0, 1, 1, 0, 0 }); // output is y2 = 0.
Console.WriteLine("output class = {0}", y2);
int y3 = hmc.Compute(new[] { 2, 2, 2, 2, 1, 1 }); // output is y3 = 1.
Console.WriteLine("output class = {0}", y3);
int y4 = hmc.Compute(new[] { 2, 2, 1, 1 }); // output is y4 = 1.
Console.WriteLine("output class = {0}", y4);
int y5 = hmc.Compute(new[] { 0, 0, 1, 3, 3, 3 }); // output is y5 = 2.
Console.WriteLine("output class = {0}", y4);
int y6 = hmc.Compute(new[] { 2, 0, 2, 2, 3, 3 }); // output is y6 = 2.
Console.WriteLine("output class = {0}", y6);
}
public void SimpleGestureRecognitionTest()
{
// Let's say we would like to do a very simple mechanism for
// gesture recognition. In this example, we will be trying to
// create a classifier that can distinguish between the words
// "hello", "car", and "wardrobe".
// Let's say we decided to acquire some data, and we asked some
// people to perform those words in front of a Kinect camera, and,
// using Microsoft's SDK, we were able to captured the x and y
// coordinates of each hand while the word was being performed.
// Let's say we decided to represent our frames as:
//
// double[] frame = { leftHandX, leftHandY, rightHandX, rightHandY };
//
// Since we captured words, this means we captured sequences of
// frames as we described above. Let's write some of those as
// rough examples to explain how gesture recognition can be done:
double[][] hello =
{
new double[] { 1.0, 0.1, 0.0, 0.0 }, // let's say the word
new double[] { 0.0, 1.0, 0.1, 0.1 }, // hello took 6 frames
new double[] { 0.0, 1.0, 0.1, 0.1 }, // to be recorded.
new double[] { 0.0, 0.0, 1.0, 0.0 },
new double[] { 0.0, 0.0, 1.0, 0.0 },
new double[] { 0.0, 0.0, 0.1, 1.1 },
};
double[][] car =
{
new double[] { 0.0, 0.0, 0.0, 1.0 }, // the car word
new double[] { 0.1, 0.0, 1.0, 0.1 }, // took only 4.
new double[] { 0.0, 0.0, 0.1, 0.0 },
new double[] { 1.0, 0.0, 0.0, 0.0 },
};
double[][] wardrobe =
{
new double[] { 0.0, 0.0, 1.0, 0.0 }, // same for the
new double[] { 0.1, 0.0, 1.0, 0.1 }, // wardrobe word.
new double[] { 0.0, 0.1, 1.0, 0.0 },
new double[] { 0.1, 0.0, 1.0, 0.1 },
};
// Here, please note that a real-world example would involve *lots*
// of samples for each word. Here, we are considering just one from
// each class which is clearly sub-optimal and should _never_ be done
// on practice. For example purposes, however, please disregard this.
// Those are the words we have in our vocabulary:
//
double[][][] words = { hello, car, wardrobe };
// Now, let's associate integer labels with them. This is needed
// for the case where there are multiple samples for each word.
//
int[] labels = { 0, 1, 2 };
// We will create our classifiers assuming an independent
// Gaussian distribution for each component in our feature
// vectors (like assuming a Naive Bayes assumption).
Independent initial = new Independent
(
new NormalDistribution(0, 1),
new NormalDistribution(0, 1),
new NormalDistribution(0, 1),
new NormalDistribution(0, 1)
);
// Now, we can proceed and create our classifier.
//
int numberOfWords = 3; // we are trying to distinguish between 3 words
int numberOfStates = 5; // this value can be found by trial-and-error
var classifier = new HiddenMarkovClassifier <Independent>
(
classes: numberOfWords,
topology: new Forward(numberOfStates), // word classifiers should use a forward topology
initial: initial
);
// Create a new learning algorithm to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning <Independent>(classifier,
// Train each model until the log-likelihood changes less than 0.001
modelIndex => new BaumWelchLearning <Independent>(classifier.Models[modelIndex])
{
Tolerance = 0.001,
Iterations = 100,
// This is necessary so the code doesn't blow up when it realize
// there is only one sample per word class. But this could also be
// needed in normal situations as well.
//
FittingOptions = new IndependentOptions()
{
InnerOption = new NormalOptions()
{
Regularization = 1e-5
}
}
}
);
// Finally, we can run the learning algorithm!
double logLikelihood = teacher.Run(words, labels);
// At this point, the classifier should be successfully
// able to distinguish between our three word classes:
//
int c1 = classifier.Compute(hello);
int c2 = classifier.Compute(car);
int c3 = classifier.Compute(wardrobe);
Assert.AreEqual(0, c1);
Assert.AreEqual(1, c2);
Assert.AreEqual(2, c3);
}
private static void check4(double[][][] words, HiddenMarkovClassifier<Independent> model, MarkovMultivariateFunction target, HiddenConditionalRandomField<double[]> hcrf)
{
double actual;
double expected;
foreach (var x in words)
{
for (int c = 0; c < model.Classes; c++)
{
for (int i = 0; i < model[c].States; i++)
{
// Check initial state transitions
double xa = model.Priors[c];
double xb = Math.Exp(model[c].Probabilities[i]);
double xc = model[c].Emissions[i].ProbabilityDensityFunction(x[0]);
expected = xa * xb * xc;
actual = Math.Exp(target.Factors[c].Compute(-1, i, x, 0, c));
Assert.IsTrue(expected.IsRelativelyEqual(actual, 1e-10));
Assert.IsFalse(double.IsNaN(actual));
}
for (int t = 1; t < x.Length; t++)
{
// Check normal state transitions
for (int i = 0; i < model[c].States; i++)
{
for (int j = 0; j < model[c].States; j++)
{
double xb = Math.Exp(model[c].Transitions[i, j]);
double xc = model[c].Emissions[j].ProbabilityDensityFunction(x[t]);
expected = xb * xc;
actual = Math.Exp(target.Factors[c].Compute(i, j, x, t, c));
Assert.IsTrue(expected.IsRelativelyEqual(actual, 1e-10));
Assert.IsFalse(double.IsNaN(actual));
}
}
}
actual = Math.Exp(model.LogLikelihood(x, c));
expected = Math.Exp(hcrf.LogLikelihood(x, c));
Assert.AreEqual(expected, actual, 1e-10);
Assert.IsFalse(double.IsNaN(actual));
actual = model.Compute(x);
expected = hcrf.Compute(x);
Assert.AreEqual(expected, actual);
Assert.IsFalse(double.IsNaN(actual));
}
}
}
public void LearnTest2()
{
// Declare some testing data
int[][] inputs = new int[][]
{
new int[] { 0,0,1,2 }, // Class 0
new int[] { 0,1,1,2 }, // Class 0
new int[] { 0,0,0,1,2 }, // Class 0
new int[] { 0,1,2,2,2 }, // Class 0
new int[] { 2,2,1,0 }, // Class 1
new int[] { 2,2,2,1,0 }, // Class 1
new int[] { 2,2,2,1,0 }, // Class 1
new int[] { 2,2,2,2,1 }, // Class 1
};
int[] outputs = new int[]
{
0,0,0,0, // First four sequences are of class 0
1,1,1,1, // Last four sequences are of class 1
};
// We are trying to predict two different classes
int classes = 2;
// Each sequence may have up to 3 symbols (0,1,2)
int symbols = 3;
// Nested models will have 3 states each
int[] states = new int[] { 3, 3 };
// Creates a new Hidden Markov Model Classifier with the given parameters
HiddenMarkovClassifier classifier = new HiddenMarkovClassifier(classes, states, symbols);
// Create a new learning algorithm to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning(classifier,
// Train each model until the log-likelihood changes less than 0.001
modelIndex => new BaumWelchLearning(classifier.Models[modelIndex])
{
Tolerance = 0.001,
Iterations = 0
}
);
// Enable support for sequence rejection
teacher.Rejection = true;
// Train the sequence classifier using the algorithm
double likelihood = teacher.Run(inputs, outputs);
HiddenMarkovModel threshold = classifier.Threshold;
Assert.AreEqual(6, threshold.States);
Assert.AreEqual(classifier.Models[0].Transitions[0, 0], threshold.Transitions[0, 0], 1e-10);
Assert.AreEqual(classifier.Models[0].Transitions[1, 1], threshold.Transitions[1, 1], 1e-10);
Assert.AreEqual(classifier.Models[0].Transitions[2, 2], threshold.Transitions[2, 2], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[0, 0], threshold.Transitions[3, 3], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[1, 1], threshold.Transitions[4, 4], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[2, 2], threshold.Transitions[5, 5], 1e-10);
for (int i = 0; i < 3; i++)
for (int j = 3; j < 6; j++)
Assert.AreEqual(Double.NegativeInfinity, threshold.Transitions[i, j]);
for (int i = 3; i < 6; i++)
for (int j = 0; j < 3; j++)
Assert.AreEqual(Double.NegativeInfinity, threshold.Transitions[i, j]);
Assert.IsFalse(Matrix.HasNaN(threshold.Transitions));
classifier.Sensitivity = 0.5;
// Will assert the models have learned the sequences correctly.
for (int i = 0; i < inputs.Length; i++)
{
int expected = outputs[i];
int actual = classifier.Compute(inputs[i], out likelihood);
Assert.AreEqual(expected, actual);
}
int[] r0 = new int[] { 1, 1, 0, 0, 2 };
double logRejection;
int c = classifier.Compute(r0, out logRejection);
Assert.AreEqual(-1, c);
Assert.AreEqual(0.99906957195279988, logRejection);
Assert.IsFalse(double.IsNaN(logRejection));
logRejection = threshold.Evaluate(r0);
Assert.AreEqual(-4.5653702970734793, logRejection, 1e-10);
Assert.IsFalse(double.IsNaN(logRejection));
threshold.Decode(r0, out logRejection);
Assert.AreEqual(-8.21169955167614, logRejection, 1e-10);
Assert.IsFalse(double.IsNaN(logRejection));
foreach (var model in classifier.Models)
{
double[,] A = model.Transitions;
for (int i = 0; i < A.GetLength(0); i++)
{
double[] row = A.Exp().GetRow(i);
double sum = row.Sum();
Assert.AreEqual(1, sum, 1e-10);
}
}
{
double[,] A = classifier.Threshold.Transitions;
for (int i = 0; i < A.GetLength(0); i++)
{
double[] row = A.GetRow(i);
double sum = row.Exp().Sum();
Assert.AreEqual(1, sum, 1e-6);
}
}
}
public static void LearnAndPredict()
{
// Suppose we would like to learn how to classify the
// following set of sequences among three class labels:
int[][] inputSequences =
{
// First class of sequences: starts and
// ends with zeros, ones in the middle:
new[] { 0, 1, 1, 1, 0 },
new[] { 0, 0, 1, 1,0, 0 },
new[] { 0, 1, 1, 1,1, 0 },
// Second class of sequences: starts with
// twos and switches to ones until the end.
new[] { 2, 2, 2, 2,1, 1, 1, 1, 1 },
new[] { 2, 2, 1, 2,1, 1, 1, 1, 1 },
new[] { 2, 2, 2, 2,2, 1, 1, 1, 1 },
// Third class of sequences: can start
// with any symbols, but ends with three.
new[] { 0, 0, 1, 1,3, 3, 3, 3 },
new[] { 0, 0, 0, 3,3, 3, 3 },
new[] { 1, 0, 1, 2,2, 2, 3, 3 },
new[] { 1, 1, 2, 3,3, 3, 3 },
new[] { 0, 0, 1, 1,3, 3, 3, 3 },
new[] { 2, 2, 0, 3,3, 3, 3 },
new[] { 1, 0, 1, 2,3, 3, 3, 3 },
new[] { 1, 1, 2, 3,3, 3, 3 },
};
// Now consider their respective class labels
int[] outputLabels =
{
/* Sequences 1-3 are from class 0: */ 0, 0, 0,
/* Sequences 4-6 are from class 1: */ 1, 1, 1,
/* Sequences 7-14 are from class 2: */ 2, 2, 2, 2, 2, 2, 2, 2
};
// We will use a single topology for all inner models, but we
// could also have explicit led different topologies for each:
ITopology forward = new Forward(state_count: 3);
// Now we create a hidden Markov classifier with the given topology
HiddenMarkovClassifier classifier = new HiddenMarkovClassifier(class_count: 3, topology: forward, symbol_count: 4);
// And create a algorithms to teach each of the inner models
var teacher = new HiddenMarkovClassifierLearning(classifier,
// We can specify individual training options for each inner model:
modelIndex => new BaumWelchLearning(classifier.Models[modelIndex])
{
Tolerance = 0.001, // iterate until log-likelihood changes less than 0.001
Iterations = 0 // don't place an upper limit on the number of iterations
});
// Then let's call its Run method to start learning
double error = teacher.Run(inputSequences, outputLabels);
// After training has finished, we can check the
// output classificaton label for some sequences.
int y1 = classifier.Compute(new[] { 0, 1, 1, 1, 0 }); // output is y1 = 0
int y2 = classifier.Compute(new[] { 0, 0, 1, 1, 0, 0 }); // output is y1 = 0
int y3 = classifier.Compute(new[] { 2, 2, 2, 2, 1, 1 }); // output is y2 = 1
int y4 = classifier.Compute(new[] { 2, 2, 1, 1 }); // output is y2 = 1
int y5 = classifier.Compute(new[] { 0, 0, 1, 3, 3, 3 }); // output is y3 = 2
int y6 = classifier.Compute(new[] { 2, 0, 2, 2, 3, 3 }); // output is y3 = 2
Console.WriteLine("y1 = {0}", y1);
Console.WriteLine("y2 = {0}", y2);
Console.WriteLine("y3 = {0}", y3);
Console.WriteLine("y4 = {0}", y4);
Console.WriteLine("y5 = {0}", y5);
Console.WriteLine("y6 = {0}", y6);
}
private void btnTrain_Click(object sender, EventArgs e)
{
if (dataGridView1.Rows.Count == 0)
{
MessageBox.Show("Please load or insert some data first.");
return;
}
int states = (int)numStates.Value;
int iterations = (int)numIterations.Value;
double tolerance = (double)numConvergence.Value;
if (rbStopIterations.Checked) tolerance = 0.0;
if (rbStopConvergence.Checked) iterations = 0;
// Retrieve the training data from the data grid view
int rows = dataGridView1.Rows.Count;
int[] outputs = new int[rows];
var sequences = new int[rows][];
for (int i = 0; i < rows; i++)
{
outputs[i] = (int)dataGridView1.Rows[i].Cells["colLabel"].Value - 1;
sequences[i] = GetFeatures((double[][])dataGridView1.Rows[i].Tag);
}
int classes = outputs.Distinct().Count();
string[] labels = new string[classes];
for (int i = 0; i < labels.Length; i++)
labels[i] = (i+1).ToString();
// Create a sequence classifier for 3 classes
classifier = new HiddenMarkovClassifier(labels.Length,
new Forward(states), symbols: 20, names: labels);
// Create the learning algorithm for the ensemble classifier
var teacher = new HiddenMarkovClassifierLearning(classifier,
// Train each model using the selected convergence criteria
i => new BaumWelchLearning(classifier.Models[i])
{
Tolerance = tolerance,
Iterations = iterations,
}
);
// Create and use a rejection threshold model
teacher.Rejection = cbRejection.Checked;
teacher.Empirical = true;
teacher.Smoothing = (double)numSmoothing.Value;
// Run the learning algorithm
teacher.Run(sequences, outputs);
double error = classifier.LogLikelihood(sequences, outputs);
int hits = 0;
toolStripProgressBar1.Visible = true;
toolStripProgressBar1.Value = 0;
toolStripProgressBar1.Step = 1;
toolStripProgressBar1.Maximum = dataGridView1.Rows.Count;
for (int i = 0; i < rows; i++)
{
double likelihood;
int index = classifier.Compute(sequences[i], out likelihood);
DataGridViewRow row = dataGridView1.Rows[i];
if (index == -1)
{
row.Cells["colClassification"].Value = String.Empty;
}
else
{
row.Cells["colClassification"].Value = classifier.Models[index].Tag;
}
int expected = (int)row.Cells["colLabel"].Value;
if (expected == index + 1)
{
row.Cells[0].Style.BackColor = Color.LightGreen;
row.Cells[1].Style.BackColor = Color.LightGreen;
row.Cells[2].Style.BackColor = Color.LightGreen;
hits++;
}
else
{
row.Cells[0].Style.BackColor = Color.White;
row.Cells[1].Style.BackColor = Color.White;
row.Cells[2].Style.BackColor = Color.White;
}
toolStripProgressBar1.PerformStep();
}
dgvModels.DataSource = classifier.Models;
toolStripProgressBar1.Visible = false;
toolStripStatusLabel1.Text = String.Format("Training complete. Hits: {0}/{1} ({2:0%})",
hits, dataGridView1.Rows.Count, (double)hits / dataGridView1.Rows.Count);
}
private static double testThresholdModel(double[][] inputs, int[] outputs, HiddenMarkovClassifier<GeneralDiscreteDistribution> classifier, double likelihood)
{
var threshold = classifier.Threshold;
Assert.AreEqual(classifier.Models[0].Transitions[0, 0], threshold.Transitions[0, 0], 1e-10);
Assert.AreEqual(classifier.Models[0].Transitions[1, 1], threshold.Transitions[1, 1], 1e-10);
Assert.AreEqual(classifier.Models[0].Transitions[2, 2], threshold.Transitions[2, 2], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[0, 0], threshold.Transitions[3, 3], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[1, 1], threshold.Transitions[4, 4], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[2, 2], threshold.Transitions[5, 5], 1e-10);
for (int i = 0; i < 3; i++)
for (int j = 3; j < 6; j++)
Assert.AreEqual(Double.NegativeInfinity, threshold.Transitions[i, j]);
for (int i = 3; i < 6; i++)
for (int j = 0; j < 3; j++)
Assert.AreEqual(Double.NegativeInfinity, threshold.Transitions[i, j]);
Assert.IsFalse(Matrix.HasNaN(threshold.Transitions));
classifier.Sensitivity = 0.5;
// Will assert the models have learned the sequences correctly.
for (int i = 0; i < inputs.Length; i++)
{
int expected = outputs[i];
int actual = classifier.Compute(inputs[i], out likelihood);
Assert.AreEqual(expected, actual);
}
double[] r0 = new double[] { 1, 1, 0, 0, 2 };
double logRejection;
int c = classifier.Compute(r0, out logRejection);
Assert.AreEqual(-1, c);
Assert.AreEqual(0.99993993054384978, logRejection);
Assert.IsFalse(double.IsNaN(logRejection));
logRejection = threshold.Evaluate(r0);
Assert.AreEqual(-5.6367018741984483, logRejection);
Assert.IsFalse(double.IsNaN(logRejection));
threshold.Decode(r0, out logRejection);
Assert.AreEqual(-8.1618027917853073, logRejection);
Assert.IsFalse(double.IsNaN(logRejection));
foreach (var model in classifier.Models)
{
double[,] A = model.Transitions;
for (int i = 0; i < A.GetLength(0); i++)
{
double[] row = A.Exp().GetRow(i);
double sum = row.Sum();
Assert.AreEqual(1, sum, 1e-10);
}
}
{
double[,] A = classifier.Threshold.Transitions;
for (int i = 0; i < A.GetLength(0); i++)
{
double[] row = A.GetRow(i);
double sum = row.Exp().Sum();
Assert.AreEqual(1, sum, 1e-6);
}
}
return likelihood;
}
private static double testThresholdModel(int[][] inputs, int[] outputs, HiddenMarkovClassifier classifier, double likelihood)
{
HiddenMarkovModel threshold = classifier.Threshold;
Assert.AreEqual(6, threshold.States);
Assert.AreEqual(classifier.Models[0].Transitions[0, 0], threshold.Transitions[0, 0], 1e-10);
Assert.AreEqual(classifier.Models[0].Transitions[1, 1], threshold.Transitions[1, 1], 1e-10);
Assert.AreEqual(classifier.Models[0].Transitions[2, 2], threshold.Transitions[2, 2], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[0, 0], threshold.Transitions[3, 3], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[1, 1], threshold.Transitions[4, 4], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[2, 2], threshold.Transitions[5, 5], 1e-10);
for (int i = 0; i < 3; i++)
for (int j = 3; j < 6; j++)
Assert.AreEqual(Double.NegativeInfinity, threshold.Transitions[i, j]);
for (int i = 3; i < 6; i++)
for (int j = 0; j < 3; j++)
Assert.AreEqual(Double.NegativeInfinity, threshold.Transitions[i, j]);
Assert.IsFalse(Matrix.HasNaN(threshold.Transitions));
classifier.Sensitivity = 0.5;
// Will assert the models have learned the sequences correctly.
for (int i = 0; i < inputs.Length; i++)
{
int expected = outputs[i];
int actual = classifier.Compute(inputs[i], out likelihood);
Assert.AreEqual(expected, actual);
}
int[] r0 = new int[] { 1, 1, 0, 0, 2 };
double logRejection;
int c = classifier.Compute(r0, out logRejection);
Assert.AreEqual(-1, c);
Assert.AreEqual(0.99994164708402866, logRejection);
Assert.IsFalse(double.IsNaN(logRejection));
logRejection = threshold.Evaluate(r0);
Assert.AreEqual(-5.6077079936209504, logRejection, 1e-10);
Assert.IsFalse(double.IsNaN(logRejection));
threshold.Decode(r0, out logRejection);
Assert.AreEqual(-9.3103554170761686, logRejection, 1e-10);
Assert.IsFalse(double.IsNaN(logRejection));
foreach (var model in classifier.Models)
{
double[,] A = model.Transitions;
for (int i = 0; i < A.GetLength(0); i++)
{
double[] row = A.Exp().GetRow(i);
double sum = row.Sum();
Assert.AreEqual(1, sum, 1e-10);
}
}
{
double[,] A = classifier.Threshold.Transitions;
for (int i = 0; i < A.GetLength(0); i++)
{
double[] row = A.GetRow(i);
double sum = row.Exp().Sum();
Assert.AreEqual(1, sum, 1e-6);
}
}
return likelihood;
}
public void LearnTest2()
{
// Declare some testing data
int[][] inputs = new int[][]
{
new int[] { 0, 0, 1, 2 }, // Class 0
new int[] { 0, 1, 1, 2 }, // Class 0
new int[] { 0, 0, 0, 1, 2 }, // Class 0
new int[] { 0, 1, 2, 2, 2 }, // Class 0
new int[] { 2, 2, 1, 0 }, // Class 1
new int[] { 2, 2, 2, 1, 0 }, // Class 1
new int[] { 2, 2, 2, 1, 0 }, // Class 1
new int[] { 2, 2, 2, 2, 1 }, // Class 1
};
int[] outputs = new int[]
{
0, 0, 0, 0, // First four sequences are of class 0
1, 1, 1, 1, // Last four sequences are of class 1
};
// We are trying to predict two different classes
int classes = 2;
// Each sequence may have up to 3 symbols (0,1,2)
int symbols = 3;
// Nested models will have 3 states each
int[] states = new int[] { 3, 3 };
// Creates a new Hidden Markov Model Classifier with the given parameters
HiddenMarkovClassifier classifier = new HiddenMarkovClassifier(classes, states, symbols);
// Create a new learning algorithm to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning(classifier,
// Train each model until the log-likelihood changes less than 0.001
modelIndex => new BaumWelchLearning(classifier.Models[modelIndex])
{
Tolerance = 0.001,
Iterations = 0
}
);
// Enable support for sequence rejection
teacher.Rejection = true;
// Train the sequence classifier using the algorithm
double likelihood = teacher.Run(inputs, outputs);
// Will assert the models have learned the sequences correctly.
for (int i = 0; i < inputs.Length; i++)
{
int expected = outputs[i];
int actual = classifier.Compute(inputs[i], out likelihood);
Assert.AreEqual(expected, actual);
}
HiddenMarkovModel threshold = classifier.Threshold;
Assert.AreEqual(6, threshold.States);
Assert.AreEqual(classifier.Models[0].Transitions[0, 0], threshold.Transitions[0, 0], 1e-10);
Assert.AreEqual(classifier.Models[0].Transitions[1, 1], threshold.Transitions[1, 1], 1e-10);
Assert.AreEqual(classifier.Models[0].Transitions[2, 2], threshold.Transitions[2, 2], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[0, 0], threshold.Transitions[3, 3], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[1, 1], threshold.Transitions[4, 4], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[2, 2], threshold.Transitions[5, 5], 1e-10);
Assert.IsFalse(Matrix.HasNaN(threshold.Transitions));
int[] r0 = new int[] { 1, 1, 0, 0, 2 };
double logRejection;
int c = classifier.Compute(r0, out logRejection);
Assert.AreEqual(-1, c);
Assert.AreEqual(0.99569011079012049, logRejection);
Assert.IsFalse(double.IsNaN(logRejection));
logRejection = threshold.Evaluate(r0);
Assert.AreEqual(-6.7949285513628528, logRejection, 1e-10);
Assert.IsFalse(double.IsNaN(logRejection));
threshold.Decode(r0, out logRejection);
Assert.AreEqual(-8.902077561009957, logRejection, 1e-10);
Assert.IsFalse(double.IsNaN(logRejection));
}
private void MainForm_MouseMove(object sender, System.Windows.Forms.MouseEventArgs e)
{
if (e.Button == MouseButtons.Left)
{
if (_isDown)
{
if (_points.Count >= 2)
{
PointR p = (PointR)_points[_points.Count - 1];
//_directionalCodewords = "" + getDirectionalCodewords(e.X, e.Y, p.X, p.Y);
_directionalCodewordsQueue.Enqueue(getDirectionalCodewords(e.X, e.Y, p.X, p.Y));
if (_directionalCodewordsQueue.Count > _maxCount)
{
_directionalCodewordsQueue.Dequeue();
}
}
_points.Add(new PointR(e.X, e.Y, Environment.TickCount));
//info = info + "a";
if (_hmmc != null && _directionalCodewordsQueue.Count > 1 && !_recording) // not recording, so testing
{
Queue <int> directionalCodewordsQueueTemp = _directionalCodewordsQueue;
while (directionalCodewordsQueueTemp.Count > 40)
{
//lblResult.Text = "Recognizing...";
_info = null;
_info = _info + _rec.encode(_directionalCodewordsQueue.ToArray()) + "\n";
//int[] observations = _rec.decode(_directionalCodewords);
int[] observations = directionalCodewordsQueueTemp.ToArray();
_info = _info + _hmmc.Compute(observations) + "\n";
string gestureName = (string)_hmms[0].Tag;
double probTemp = 0;
_hmmc[0].Decode(observations, out probTemp);
//double probTemp = hmms[0].Evaluate(observations);
foreach (HiddenMarkovModel hmm in _hmms)
{
//double prob = hmm.Evaluate(observations);
double prob = 0;
int[] viterbipath = hmm.Decode(observations, out prob);
if (prob > probTemp)
{
gestureName = (string)hmm.Tag;
probTemp = prob;
}
//info = info + hmm.Tag + "\t" + hmm.Evaluate(observations) + "\t";
_info = _info + hmm.Tag + "\t" + prob + "\t";
// = hmm.Decode(observations);
foreach (int state in viterbipath)
{
_info = _info + state + " ";
}
_info = _info + "\n";
}
double probTM = 0;
int[] viterbipathTM = _hmmc.Threshold.Decode(observations, out probTM);
_info = _info + "ThresholdModel\t" + probTM + "\t";
//hmmc.Threshold.Decode(observations);
foreach (int state in viterbipathTM)
{
_info = _info + state + " ";
}
_info = _info + "\n";
if (probTM > probTemp)
{
gestureName = "Threshold";
_info = _info + "\n\n" + gestureName;
}
else
{
_info = _info + "\n\n" + gestureName;
_directionalCodewordsQueue.Clear();
break;
}
for (int loop = 0; loop < 10; loop++)
{
directionalCodewordsQueueTemp.Dequeue();
}
}
}
Invalidate();
}
}
//Invalidate(new Rectangle(e.X - 2, e.Y - 2, 4, 4));
}
private void btnLearnHMM_Click(object sender, EventArgs e)
{
if (gridSamples.Rows.Count == 0)
{
MessageBox.Show("Please load or insert some data first.");
return;
}
BindingList <Sequence> samples = database.Samples;
BindingList <String> classes = database.Classes;
double[][][] inputs = new double[samples.Count][][];
int[] outputs = new int[samples.Count];
for (int i = 0; i < inputs.Length; i++)
{
inputs[i] = samples[i].Input;
outputs[i] = samples[i].Output;
}
int states = 5;
int iterations = 0;
double tolerance = 0.01;
bool rejection = false;
hmm = new HiddenMarkovClassifier <MultivariateNormalDistribution>(classes.Count,
new Forward(states), new MultivariateNormalDistribution(2), classes.ToArray());
// Create the learning algorithm for the ensemble classifier
var teacher = new HiddenMarkovClassifierLearning <MultivariateNormalDistribution>(hmm,
// Train each model using the selected convergence criteria
i => new BaumWelchLearning <MultivariateNormalDistribution>(hmm.Models[i])
{
Tolerance = tolerance,
Iterations = iterations,
FittingOptions = new NormalOptions()
{
Regularization = 1e-5
}
}
);
teacher.Empirical = true;
teacher.Rejection = rejection;
// Run the learning algorithm
double error = teacher.Run(inputs, outputs);
// Classify all training instances
foreach (var sample in database.Samples)
{
sample.RecognizedAs = hmm.Compute(sample.Input);
}
foreach (DataGridViewRow row in gridSamples.Rows)
{
var sample = row.DataBoundItem as Sequence;
row.DefaultCellStyle.BackColor = (sample.RecognizedAs == sample.Output) ?
Color.LightGreen : Color.White;
}
btnLearnHCRF.Enabled = true;
hcrf = null;
}
public void LearnTest6()
{
// Create a Continuous density Hidden Markov Model Sequence Classifier
// to detect a multivariate sequence and the same sequence backwards.
double[][][] sequences = new double[][][]
{
new double[][]
{
// This is the first sequence with label = 0
new double[] { 0, 1 },
new double[] { 1, 2 },
new double[] { 2, 3 },
new double[] { 3, 4 },
new double[] { 4, 5 },
},
new double[][]
{
// This is the second sequence with label = 1
new double[] { 4, 3 },
new double[] { 3, 2 },
new double[] { 2, 1 },
new double[] { 1, 0 },
new double[] { 0, -1 },
}
};
// Labels for the sequences
int[] labels = { 0, 1 };
var density = new MultivariateNormalDistribution(2);
// Creates a sequence classifier containing 2 hidden Markov Models with 2 states
// and an underlying multivariate mixture of Normal distributions as density.
var classifier = new HiddenMarkovClassifier <MultivariateNormalDistribution>(
2, new Custom(new double[2, 2], new double[2]), density);
// Configure the learning algorithms to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning <MultivariateNormalDistribution>(
classifier,
// Train each model until the log-likelihood changes less than 0.0001
modelIndex => new BaumWelchLearning <MultivariateNormalDistribution>(
classifier.Models[modelIndex])
{
Tolerance = 0.0001,
Iterations = 0,
FittingOptions = new NormalOptions()
{
Diagonal = true
}
}
);
// Train the sequence classifier using the algorithm
double logLikelihood = teacher.Run(sequences, labels);
// Calculate the probability that the given
// sequences originated from the model
double response1, response2;
// Try to classify the 1st sequence (output should be 0)
int c1 = classifier.Compute(sequences[0], out response1);
// Try to classify the 2nd sequence (output should be 1)
int c2 = classifier.Compute(sequences[1], out response2);
Assert.AreEqual(double.NegativeInfinity, logLikelihood);
Assert.AreEqual(0, response1);
Assert.AreEqual(0, response2);
Assert.IsFalse(double.IsNaN(logLikelihood));
Assert.IsFalse(double.IsNaN(response1));
Assert.IsFalse(double.IsNaN(response2));
}
private static double testThresholdModel(int[][] inputs, int[] outputs, HiddenMarkovClassifier classifier, double likelihood)
{
HiddenMarkovModel threshold = classifier.Threshold;
Assert.AreEqual(6, threshold.States);
Assert.AreEqual(classifier.Models[0].Transitions[0, 0], threshold.Transitions[0, 0], 1e-10);
Assert.AreEqual(classifier.Models[0].Transitions[1, 1], threshold.Transitions[1, 1], 1e-10);
Assert.AreEqual(classifier.Models[0].Transitions[2, 2], threshold.Transitions[2, 2], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[0, 0], threshold.Transitions[3, 3], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[1, 1], threshold.Transitions[4, 4], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[2, 2], threshold.Transitions[5, 5], 1e-10);
for (int i = 0; i < 3; i++)
{
for (int j = 3; j < 6; j++)
{
Assert.AreEqual(Double.NegativeInfinity, threshold.Transitions[i, j]);
}
}
for (int i = 3; i < 6; i++)
{
for (int j = 0; j < 3; j++)
{
Assert.AreEqual(Double.NegativeInfinity, threshold.Transitions[i, j]);
}
}
Assert.IsFalse(Matrix.HasNaN(threshold.LogTransitions));
classifier.Sensitivity = 0.5;
// Will assert the models have learned the sequences correctly.
for (int i = 0; i < inputs.Length; i++)
{
int expected = outputs[i];
int actual = classifier.Compute(inputs[i], out likelihood);
Assert.AreEqual(expected, actual);
}
int[] r0 = new int[] { 1, 1, 0, 0, 2 };
double logRejection;
int c = classifier.Compute(r0, out logRejection);
Assert.AreEqual(-1, c);
Assert.AreEqual(0.99996241769427985, logRejection, 1e-10);
logRejection = threshold.Evaluate(r0);
Assert.AreEqual(-5.5993214137039073, logRejection, 1e-10);
threshold.Decode(r0, out logRejection);
Assert.AreEqual(-9.31035541707617, logRejection, 1e-10);
foreach (var model in classifier.Models)
{
double[,] A = model.Transitions;
for (int i = 0; i < A.GetLength(0); i++)
{
double[] row = A.Exp().GetRow(i);
double sum = row.Sum();
Assert.AreEqual(1, sum, 1e-10);
}
}
{
double[,] A = classifier.Threshold.Transitions;
for (int i = 0; i < A.GetLength(0); i++)
{
double[] row = A.GetRow(i);
double sum = row.Exp().Sum();
Assert.AreEqual(1, sum, 1e-6);
}
}
return(likelihood);
}
public void LearnTest7()
{
// Create a Continuous density Hidden Markov Model Sequence Classifier
// to detect a multivariate sequence and the same sequence backwards.
double[][][] sequences = new double[][][]
{
new double[][]
{
// This is the first sequence with label = 0
new double[] { 0, 1 },
new double[] { 1, 2 },
new double[] { 2, 3 },
new double[] { 3, 4 },
new double[] { 4, 5 },
},
new double[][]
{
// This is the second sequence with label = 1
new double[] { 4, 3 },
new double[] { 3, 2 },
new double[] { 2, 1 },
new double[] { 1, 0 },
new double[] { 0, -1 },
}
};
// Labels for the sequences
int[] labels = { 0, 1 };
var initialDensity = new MultivariateNormalDistribution(2);
// Creates a sequence classifier containing 2 hidden Markov Models with 2 states
// and an underlying multivariate mixture of Normal distributions as density.
var classifier = new HiddenMarkovClassifier<MultivariateNormalDistribution>(
classes: 2, topology: new Forward(2), initial: initialDensity);
// Configure the learning algorithms to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning<MultivariateNormalDistribution>(
classifier,
// Train each model until the log-likelihood changes less than 0.0001
modelIndex => new BaumWelchLearning<MultivariateNormalDistribution>(
classifier.Models[modelIndex])
{
Tolerance = 0.0001,
Iterations = 0,
FittingOptions = new NormalOptions()
{
Diagonal = true, // only diagonal covariance matrices
Regularization = 1e-5 // avoid non-positive definite errors
}
}
);
// Train the sequence classifier using the algorithm
double logLikelihood = teacher.Run(sequences, labels);
// Calculate the probability that the given
// sequences originated from the model
double likelihood, likelihood2;
// Try to classify the 1st sequence (output should be 0)
int c1 = classifier.Compute(sequences[0], out likelihood);
// Try to classify the 2nd sequence (output should be 1)
int c2 = classifier.Compute(sequences[1], out likelihood2);
Assert.AreEqual(0, c1);
Assert.AreEqual(1, c2);
Assert.AreEqual(-24.560663315259973, logLikelihood, 1e-10);
Assert.AreEqual(0.99999999998805045, likelihood, 1e-10);
Assert.AreEqual(0.99999999998805045, likelihood2, 1e-10);
Assert.IsFalse(double.IsNaN(logLikelihood));
Assert.IsFalse(double.IsNaN(likelihood));
Assert.IsFalse(double.IsNaN(likelihood2));
}
public void SimpleGestureRecognitionTest()
{
// Let's say we would like to do a very simple mechanism for
// gesture recognition. In this example, we will be trying to
// create a classifier that can distinguish between the words
// "hello", "car", and "wardrobe".
// Let's say we decided to acquire some data, and we asked some
// people to perform those words in front of a Kinect camera, and,
// using Microsoft's SDK, we were able to captured the x and y
// coordinates of each hand while the word was being performed.
// Let's say we decided to represent our frames as:
//
// double[] frame = { leftHandX, leftHandY, rightHandX, rightHandY };
//
// Since we captured words, this means we captured sequences of
// frames as we described above. Let's write some of those as
// rough examples to explain how gesture recognition can be done:
double[][] hello =
{
new double[] { 1.0, 0.1, 0.0, 0.0 }, // let's say the word
new double[] { 0.0, 1.0, 0.1, 0.1 }, // hello took 6 frames
new double[] { 0.0, 1.0, 0.1, 0.1 }, // to be recorded.
new double[] { 0.0, 0.0, 1.0, 0.0 },
new double[] { 0.0, 0.0, 1.0, 0.0 },
new double[] { 0.0, 0.0, 0.1, 1.1 },
};
double[][] car =
{
new double[] { 0.0, 0.0, 0.0, 1.0 }, // the car word
new double[] { 0.1, 0.0, 1.0, 0.1 }, // took only 4.
new double[] { 0.0, 0.0, 0.1, 0.0 },
new double[] { 1.0, 0.0, 0.0, 0.0 },
};
double[][] wardrobe =
{
new double[] { 0.0, 0.0, 1.0, 0.0 }, // same for the
new double[] { 0.1, 0.0, 1.0, 0.1 }, // wardrobe word.
new double[] { 0.0, 0.1, 1.0, 0.0 },
new double[] { 0.1, 0.0, 1.0, 0.1 },
};
// Here, please note that a real-world example would involve *lots*
// of samples for each word. Here, we are considering just one from
// each class which is clearly sub-optimal and should _never_ be done
// on practice. For example purposes, however, please disregard this.
// Those are the words we have in our vocabulary:
//
double[][][] words = { hello, car, wardrobe };
// Now, let's associate integer labels with them. This is needed
// for the case where there are multiple samples for each word.
//
int[] labels = { 0, 1, 2 };
// We will create our classifiers assuming an independent
// Gaussian distribution for each component in our feature
// vectors (like assuming a Naive Bayes assumption).
Independent initial = new Independent
(
new NormalDistribution(0, 1),
new NormalDistribution(0, 1),
new NormalDistribution(0, 1),
new NormalDistribution(0, 1)
);
// Now, we can proceed and create our classifier.
//
int numberOfWords = 3; // we are trying to distinguish between 3 words
int numberOfStates = 5; // this value can be found by trial-and-error
var classifier = new HiddenMarkovClassifier<Independent>
(
classes: numberOfWords,
topology: new Forward(numberOfStates), // word classifiers should use a forward topology
initial: initial
);
// Create a new learning algorithm to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning<Independent>(classifier,
// Train each model until the log-likelihood changes less than 0.001
modelIndex => new BaumWelchLearning<Independent>(classifier.Models[modelIndex])
{
Tolerance = 0.001,
Iterations = 100,
// This is necessary so the code doesn't blow up when it realize
// there is only one sample per word class. But this could also be
// needed in normal situations as well.
//
FittingOptions = new IndependentOptions()
{
InnerOption = new NormalOptions() { Regularization = 1e-5 }
}
}
);
// Finally, we can run the learning algorithm!
double logLikelihood = teacher.Run(words, labels);
// At this point, the classifier should be successfully
// able to distinguish between our three word classes:
//
int c1 = classifier.Compute(hello);
int c2 = classifier.Compute(car);
int c3 = classifier.Compute(wardrobe);
Assert.AreEqual(0, c1);
Assert.AreEqual(1, c2);
Assert.AreEqual(2, c3);
}
public void LearnTest9()
{
double[][][] inputs = large_gestures;
int[] outputs = large_outputs;
int states = 5;
int iterations = 100;
double tolerance = 0.01;
bool rejection = true;
double sensitivity = 1E-85;
int dimension = inputs[0][0].Length;
var hmm = new HiddenMarkovClassifier<MultivariateNormalDistribution>(2,
new Forward(states), new MultivariateNormalDistribution(dimension));
// Create the learning algorithm for the ensemble classifier
var teacher = new HiddenMarkovClassifierLearning<MultivariateNormalDistribution>(hmm,
// Train each model using the selected convergence criteria
i => new BaumWelchLearning<MultivariateNormalDistribution>(hmm.Models[i])
{
Tolerance = tolerance,
Iterations = iterations,
FittingOptions = new NormalOptions()
{
Regularization = 1e-5
}
}
);
teacher.Empirical = true;
teacher.Rejection = rejection;
// Run the learning algorithm
double logLikelihood = teacher.Run(inputs, outputs);
hmm.Sensitivity = sensitivity;
for (int i = 0; i < large_gestures.Length; i++)
{
int actual = hmm.Compute(large_gestures[i]);
int expected = large_outputs[i];
Assert.AreEqual(expected,actual);
}
}
static void runDiscreteDensityHiddenMarkovClassifierLearningExample()
{
// Observation sequences should only contain symbols that are greater than or equal to 0, and lesser than the number of symbols.
int[][] observationSequences =
{
// First class of sequences: starts and ends with zeros, ones in the middle.
new[] { 0, 1, 1, 1, 0 },
new[] { 0, 0, 1, 1,0, 0 },
new[] { 0, 1, 1, 1,1, 0 },
// Second class of sequences: starts with twos and switches to ones until the end.
new[] { 2, 2, 2, 2,1, 1, 1, 1, 1 },
new[] { 2, 2, 1, 2,1, 1, 1, 1, 1 },
new[] { 2, 2, 2, 2,2, 1, 1, 1, 1 },
// Third class of sequences: can start with any symbols, but ends with three.
new[] { 0, 0, 1, 1,3, 3, 3, 3 },
new[] { 0, 0, 0, 3,3, 3, 3 },
new[] { 1, 0, 1, 2,2, 2, 3, 3 },
new[] { 1, 1, 2, 3,3, 3, 3 },
new[] { 0, 0, 1, 1,3, 3, 3, 3 },
new[] { 2, 2, 0, 3,3, 3, 3 },
new[] { 1, 0, 1, 2,3, 3, 3, 3 },
new[] { 1, 1, 2, 3,3, 3, 3 },
};
// Consider their respective class labels.
// Class labels have to be zero-based and successive integers.
int[] classLabels =
{
0, 0, 0, // Sequences 1-3 are from class 0.
1, 1, 1, // Sequences 4-6 are from class 1.
2, 2, 2, 2, 2, 2, 2, 2 // Sequences 7-14 are from class 2.
};
// Use a single topology for all inner models.
ITopology forward = new Forward(states: 3);
// Create a hidden Markov classifier with the given topology.
HiddenMarkovClassifier hmc = new HiddenMarkovClassifier(classes: 3, topology: forward, symbols: 4);
// Create a algorithms to teach each of the inner models.
var trainer = new HiddenMarkovClassifierLearning(
hmc,
// Specify individual training options for each inner model.
modelIndex => new BaumWelchLearning(hmc.Models[modelIndex])
{
Tolerance = 0.001, // iterate until log-likelihood changes less than 0.001.
Iterations = 0 // don't place an upper limit on the number of iterations.
}
);
// Call its Run method to start learning.
double averageLogLikelihood = trainer.Run(observationSequences, classLabels);
Console.WriteLine("average log-likelihood for the observations = {0}", averageLogLikelihood);
// Check the output classificaton label for some sequences.
int y1 = hmc.Compute(new[] { 0, 1, 1, 1, 0 }); // output is y1 = 0.
Console.WriteLine("output class = {0}", y1);
int y2 = hmc.Compute(new[] { 0, 0, 1, 1, 0, 0 }); // output is y2 = 0.
Console.WriteLine("output class = {0}", y2);
int y3 = hmc.Compute(new[] { 2, 2, 2, 2, 1, 1 }); // output is y3 = 1.
Console.WriteLine("output class = {0}", y3);
int y4 = hmc.Compute(new[] { 2, 2, 1, 1 }); // output is y4 = 1.
Console.WriteLine("output class = {0}", y4);
int y5 = hmc.Compute(new[] { 0, 0, 1, 3, 3, 3 }); // output is y5 = 2.
Console.WriteLine("output class = {0}", y4);
int y6 = hmc.Compute(new[] { 2, 0, 2, 2, 3, 3 }); // output is y6 = 2.
Console.WriteLine("output class = {0}", y6);
}
public void LearnTest2()
{
// Declare some testing data
int[][] inputs = new int[][]
{
new int[] { 0,0,1,2 }, // Class 0
new int[] { 0,1,1,2 }, // Class 0
new int[] { 0,0,0,1,2 }, // Class 0
new int[] { 0,1,2,2,2 }, // Class 0
new int[] { 2,2,1,0 }, // Class 1
new int[] { 2,2,2,1,0 }, // Class 1
new int[] { 2,2,2,1,0 }, // Class 1
new int[] { 2,2,2,2,1 }, // Class 1
};
int[] outputs = new int[]
{
0,0,0,0, // First four sequences are of class 0
1,1,1,1, // Last four sequences are of class 1
};
// We are trying to predict two different classes
int classes = 2;
// Each sequence may have up to 3 symbols (0,1,2)
int symbols = 3;
// Nested models will have 3 states each
int[] states = new int[] { 3, 3 };
// Creates a new Hidden Markov Model Classifier with the given parameters
HiddenMarkovClassifier classifier = new HiddenMarkovClassifier(classes, states, symbols);
// Create a new learning algorithm to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning(classifier,
// Train each model until the log-likelihood changes less than 0.001
modelIndex => new BaumWelchLearning(classifier.Models[modelIndex])
{
Tolerance = 0.001,
Iterations = 0
}
);
// Enable support for sequence rejection
teacher.Rejection = true;
// Train the sequence classifier using the algorithm
double likelihood = teacher.Run(inputs, outputs);
HiddenMarkovModel threshold = classifier.Threshold;
Assert.AreEqual(6, threshold.States);
Assert.AreEqual(classifier.Models[0].Transitions[0, 0], threshold.Transitions[0, 0], 1e-10);
Assert.AreEqual(classifier.Models[0].Transitions[1, 1], threshold.Transitions[1, 1], 1e-10);
Assert.AreEqual(classifier.Models[0].Transitions[2, 2], threshold.Transitions[2, 2], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[0, 0], threshold.Transitions[3, 3], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[1, 1], threshold.Transitions[4, 4], 1e-10);
Assert.AreEqual(classifier.Models[1].Transitions[2, 2], threshold.Transitions[5, 5], 1e-10);
for (int i = 0; i < 3; i++)
for (int j = 3; j < 6; j++)
Assert.AreEqual(Double.NegativeInfinity, threshold.Transitions[i, j]);
for (int i = 3; i < 6; i++)
for (int j = 0; j < 3; j++)
Assert.AreEqual(Double.NegativeInfinity, threshold.Transitions[i, j]);
Assert.IsFalse(Matrix.HasNaN(threshold.Transitions));
classifier.Sensitivity = 0.5;
// Will assert the models have learned the sequences correctly.
for (int i = 0; i < inputs.Length; i++)
{
int expected = outputs[i];
int actual = classifier.Compute(inputs[i], out likelihood);
Assert.AreEqual(expected, actual);
}
int[] r0 = new int[] { 1, 1, 0, 0, 2 };
double logRejection;
int c = classifier.Compute(r0, out logRejection);
Assert.AreEqual(-1, c);
Assert.AreEqual(0.99906957195279988, logRejection);
Assert.IsFalse(double.IsNaN(logRejection));
logRejection = threshold.Evaluate(r0);
Assert.AreEqual(-4.5653702970734793, logRejection, 1e-10);
Assert.IsFalse(double.IsNaN(logRejection));
threshold.Decode(r0, out logRejection);
Assert.AreEqual(-8.21169955167614, logRejection, 1e-10);
Assert.IsFalse(double.IsNaN(logRejection));
foreach (var model in classifier.Models)
{
double[,] A = model.Transitions;
for (int i = 0; i < A.GetLength(0); i++)
{
double[] row = A.Exp().GetRow(i);
double sum = row.Sum();
Assert.AreEqual(1, sum, 1e-10);
}
}
{
double[,] A = classifier.Threshold.Transitions;
for (int i = 0; i < A.GetLength(0); i++)
{
double[] row = A.GetRow(i);
double sum = row.Exp().Sum();
Assert.AreEqual(1, sum, 1e-6);
}
}
}
public void LearnTest()
{
// Declare some testing data
int[][] inputs = new int[][]
{
new int[] { 0,1,1,0 }, // Class 0
new int[] { 0,0,1,0 }, // Class 0
new int[] { 0,1,1,1,0 }, // Class 0
new int[] { 0,1,0 }, // Class 0
new int[] { 1,0,0,1 }, // Class 1
new int[] { 1,1,0,1 }, // Class 1
new int[] { 1,0,0,0,1 }, // Class 1
new int[] { 1,0,1 }, // Class 1
};
int[] outputs = new int[]
{
0,0,0,0, // First four sequences are of class 0
1,1,1,1, // Last four sequences are of class 1
};
// We are trying to predict two different classes
int classes = 2;
// Each sequence may have up to two symbols (0 or 1)
int symbols = 2;
// Nested models will have two states each
int[] states = new int[] { 2, 2 };
// Creates a new Hidden Markov Model Classifier with the given parameters
HiddenMarkovClassifier classifier = new HiddenMarkovClassifier(classes, states, symbols);
// Create a new learning algorithm to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning(classifier,
// Train each model until the log-likelihood changes less than 0.001
modelIndex => new BaumWelchLearning(classifier.Models[modelIndex])
{
Tolerance = 0.001,
Iterations = 0
}
);
// Train the sequence classifier using the algorithm
double likelihood = teacher.Run(inputs, outputs);
// Will assert the models have learned the sequences correctly.
for (int i = 0; i < inputs.Length; i++)
{
int expected = outputs[i];
int actual = classifier.Compute(inputs[i], out likelihood);
Assert.AreEqual(expected, actual);
}
}
public void LearnTest4()
{
// Create a Continuous density Hidden Markov Model Sequence Classifier
// to detect a multivariate sequence and the same sequence backwards.
double[][][] sequences = new double[][][]
{
new double[][]
{
// This is the first sequence with label = 0
new double[] { 0 },
new double[] { 1 },
new double[] { 2 },
new double[] { 3 },
new double[] { 4 },
},
new double[][]
{
// This is the second sequence with label = 1
new double[] { 4 },
new double[] { 3 },
new double[] { 2 },
new double[] { 1 },
new double[] { 0 },
}
};
// Labels for the sequences
int[] labels = { 0, 1 };
// Create a mixture of two 1-dimensional normal distributions (by default,
// initialized with zero mean and unit covariance matrices).
var density = new MultivariateMixture <MultivariateNormalDistribution>(
new MultivariateNormalDistribution(1),
new MultivariateNormalDistribution(1));
// Creates a sequence classifier containing 2 hidden Markov Models with 2 states
// and an underlying multivariate mixture of Normal distributions as density.
var classifier = new HiddenMarkovClassifier <MultivariateMixture <MultivariateNormalDistribution> >(
2, new Ergodic(2), density);
// Configure the learning algorithms to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning <MultivariateMixture <MultivariateNormalDistribution> >(
classifier,
// Train each model until the log-likelihood changes less than 0.0001
modelIndex => new BaumWelchLearning <MultivariateMixture <MultivariateNormalDistribution> >(
classifier.Models[modelIndex])
{
Tolerance = 0.0001,
Iterations = 0,
}
);
// Train the sequence classifier using the algorithm
double logLikelihood = teacher.Run(sequences, labels);
// Calculate the probability that the given
// sequences originated from the model
double logLikelihood1, logLikelihood2;
// Try to classify the 1st sequence (output should be 0)
int c1 = classifier.Compute(sequences[0], out logLikelihood1);
// Try to classify the 2nd sequence (output should be 1)
int c2 = classifier.Compute(sequences[1], out logLikelihood2);
Assert.AreEqual(0, c1);
Assert.AreEqual(1, c2);
Assert.AreEqual(-13.271981026832933, logLikelihood, 1e-10);
Assert.AreEqual(0.99999791320102149, logLikelihood1, 1e-10);
Assert.AreEqual(0.99999791320102149, logLikelihood2, 1e-10);
Assert.IsFalse(double.IsNaN(logLikelihood));
Assert.IsFalse(double.IsNaN(logLikelihood1));
Assert.IsFalse(double.IsNaN(logLikelihood2));
}
public void RecognizeAsync(ISoundSignalReader signal, HiddenMarkovClassifier hmm, Action<string> handleMessage,
SignalVisitor voiceVisitor = null)
{
Action<List<double[]>> action = features =>
{
var observations = _codeBook.Quantize(features.Select(item => new Point(item)).ToArray());
double[] responsabilities;
var ret = hmm.Compute(observations, out responsabilities);
var models = hmm.Models;
var likelyHoodValue = Double.MinValue;
foreach (var model in models)
{
var val = model.Evaluate(observations);
if (val > likelyHoodValue)
{
likelyHoodValue = val;
}
}
handleMessage(hmm[ret].Tag.ToString());
};
var featureUtility = new FeatureUtility(_engineParameters);
featureUtility.ExtractFeaturesAsync(signal, action, voiceVisitor);
}
private void btnTrain_Click(object sender, EventArgs e)
{
if (dataGridView1.Rows.Count == 0)
{
MessageBox.Show("Please load or insert some data first.");
return;
}
int states = (int)numStates.Value;
int iterations = (int)numIterations.Value;
double tolerance = (double)numConvergence.Value;
if (rbStopIterations.Checked)
{
tolerance = 0.0;
}
if (rbStopConvergence.Checked)
{
iterations = 0;
}
// Retrieve the training data from the data grid view
int rows = dataGridView1.Rows.Count;
int[] outputs = new int[rows];
var sequences = new int[rows][];
for (int i = 0; i < rows; i++)
{
outputs[i] = (int)dataGridView1.Rows[i].Cells["colLabel"].Value - 1;
sequences[i] = GetFeatures((double[][])dataGridView1.Rows[i].Tag);
}
int classes = outputs.Distinct().Count();
string[] labels = new string[classes];
for (int i = 0; i < labels.Length; i++)
{
labels[i] = (i + 1).ToString();
}
// Create a sequence classifier for 3 classes
classifier = new HiddenMarkovClassifier(labels.Length,
new Forward(states), symbols: 20, names: labels);
// Create the learning algorithm for the ensemble classifier
var teacher = new HiddenMarkovClassifierLearning(classifier,
// Train each model using the selected convergence criteria
i => new BaumWelchLearning(classifier.Models[i])
{
Tolerance = tolerance,
Iterations = iterations,
}
);
// Create and use a rejection threshold model
teacher.Rejection = cbRejection.Checked;
teacher.Empirical = true;
teacher.Smoothing = (double)numSmoothing.Value;
// Run the learning algorithm
teacher.Run(sequences, outputs);
double error = classifier.LogLikelihood(sequences, outputs);
int hits = 0;
toolStripProgressBar1.Visible = true;
toolStripProgressBar1.Value = 0;
toolStripProgressBar1.Step = 1;
toolStripProgressBar1.Maximum = dataGridView1.Rows.Count;
for (int i = 0; i < rows; i++)
{
double likelihood;
int index = classifier.Compute(sequences[i], out likelihood);
DataGridViewRow row = dataGridView1.Rows[i];
if (index == -1)
{
row.Cells["colClassification"].Value = String.Empty;
}
else
{
row.Cells["colClassification"].Value = classifier.Models[index].Tag;
}
int expected = (int)row.Cells["colLabel"].Value;
if (expected == index + 1)
{
row.Cells[0].Style.BackColor = Color.LightGreen;
row.Cells[1].Style.BackColor = Color.LightGreen;
row.Cells[2].Style.BackColor = Color.LightGreen;
hits++;
}
else
{
row.Cells[0].Style.BackColor = Color.White;
row.Cells[1].Style.BackColor = Color.White;
row.Cells[2].Style.BackColor = Color.White;
}
toolStripProgressBar1.PerformStep();
}
dgvModels.DataSource = classifier.Models;
toolStripProgressBar1.Visible = false;
toolStripStatusLabel1.Text = String.Format("Training complete. Hits: {0}/{1} ({2:0%})",
hits, dataGridView1.Rows.Count, (double)hits / dataGridView1.Rows.Count);
}
public void SimpleGestureRecognitionTest()
{
// Let's say we would like to do a very simple mechanism for
// gesture recognition. In this example, we will be trying to
// create a classifier that can distinguish between the words
// "hello", "car", and "wardrobe".
// Let's say we decided to acquire some data, and we asked some
// people to perform those words in front of a Kinect camera, and,
// using Microsoft's SDK, we were able to captured the x and y
// coordinates of each hand while the word was being performed.
// Let's say we decided to represent our frames as:
//
// double[] frame = { leftHandX, leftHandY, rightHandX, rightHandY };
//
// Since we captured words, this means we captured sequences of
// frames as we described above. Let's write some of those as
// rough examples to explain how gesture recognition can be done:
double[][] hello =
{
new double[] { 1.0, 0.1, 0.0, 0.0 }, // let's say the word
new double[] { 0.0, 1.0, 0.1, 0.1 }, // hello took 6 frames
new double[] { 0.0, 1.0, 0.1, 0.1 }, // to be recorded.
new double[] { 0.0, 0.0, 1.0, 0.0 },
new double[] { 0.0, 0.0, 1.0, 0.0 },
new double[] { 0.0, 0.0, 0.1, 1.1 },
};
double[][] car =
{
new double[] { 0.0, 0.0, 0.0, 1.0 }, // the car word
new double[] { 0.1, 0.0, 1.0, 0.1 }, // took only 4.
new double[] { 0.0, 0.0, 0.1, 0.0 },
new double[] { 1.0, 0.0, 0.0, 0.0 },
};
double[][] wardrobe =
{
new double[] { 0.0, 0.0, 1.0, 0.0 }, // same for the
new double[] { 0.1, 0.0, 1.0, 0.1 }, // wardrobe word.
new double[] { 0.0, 0.1, 1.0, 0.0 },
new double[] { 0.1, 0.0, 1.0, 0.1 },
};
// Here, please note that a real-world example would involve *lots*
// of samples for each word. Here, we are considering just one from
// each class which is clearly sub-optimal and should _never_ be done
// on practice. For example purposes, however, please disregard this.
// Those are the words we have in our vocabulary:
//
double[][][] words = { hello, car, wardrobe };
// Now, let's associate integer labels with them. This is needed
// for the case where there are multiple samples for each word.
//
int[] labels = { 0, 1, 2 };
// We will create our classifiers assuming an independent
// Gaussian distribution for each component in our feature
// vectors (like assuming a Naive Bayes assumption).
var initial = new Independent <NormalDistribution>
(
new NormalDistribution(0, 1),
new NormalDistribution(0, 1),
new NormalDistribution(0, 1),
new NormalDistribution(0, 1)
);
// Now, we can proceed and create our classifier.
//
int numberOfWords = 3; // we are trying to distinguish between 3 words
int numberOfStates = 5; // this value can be found by trial-and-error
var hmm = new HiddenMarkovClassifier <Independent <NormalDistribution> >
(
classes: numberOfWords,
topology: new Forward(numberOfStates), // word classifiers should use a forward topology
initial: initial
);
// Create a new learning algorithm to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning <Independent <NormalDistribution> >(hmm,
// Train each model until the log-likelihood changes less than 0.001
modelIndex => new BaumWelchLearning <Independent <NormalDistribution> >(hmm.Models[modelIndex])
{
Tolerance = 0.001,
Iterations = 100,
// This is necessary so the code doesn't blow up when it realize
// there is only one sample per word class. But this could also be
// needed in normal situations as well.
//
FittingOptions = new IndependentOptions()
{
InnerOption = new NormalOptions()
{
Regularization = 1e-5
}
}
}
);
// Finally, we can run the learning algorithm!
double logLikelihood = teacher.Run(words, labels);
// At this point, the classifier should be successfully
// able to distinguish between our three word classes:
//
int tc1 = hmm.Compute(hello);
int tc2 = hmm.Compute(car);
int tc3 = hmm.Compute(wardrobe);
Assert.AreEqual(0, tc1);
Assert.AreEqual(1, tc2);
Assert.AreEqual(2, tc3);
// Now, we can use the Markov classifier to initialize a HCRF
var function = new MarkovMultivariateFunction(hmm);
var hcrf = new HiddenConditionalRandomField <double[]>(function);
// We can check that both are equivalent, although they have
// formulations that can be learned with different methods
//
for (int i = 0; i < words.Length; i++)
{
// Should be the same
int expected = hmm.Compute(words[i]);
int actual = hcrf.Compute(words[i]);
// Should be the same
double h0 = hmm.LogLikelihood(words[i], 0);
double c0 = hcrf.LogLikelihood(words[i], 0);
double h1 = hmm.LogLikelihood(words[i], 1);
double c1 = hcrf.LogLikelihood(words[i], 1);
double h2 = hmm.LogLikelihood(words[i], 2);
double c2 = hcrf.LogLikelihood(words[i], 2);
Assert.AreEqual(expected, actual);
Assert.AreEqual(h0, c0, 1e-10);
Assert.IsTrue(h1.IsRelativelyEqual(c1, 1e-10));
Assert.IsTrue(h2.IsRelativelyEqual(c2, 1e-10));
Assert.IsFalse(double.IsNaN(c0));
Assert.IsFalse(double.IsNaN(c1));
Assert.IsFalse(double.IsNaN(c2));
}
// Now we can learn the HCRF using one of the best learning
// algorithms available, Resilient Backpropagation learning:
// Create a learning algorithm
var rprop = new HiddenResilientGradientLearning <double[]>(hcrf)
{
Iterations = 50,
Tolerance = 1e-5
};
// Run the algorithm and learn the models
double error = rprop.Run(words, labels);
// At this point, the HCRF should be successfully
// able to distinguish between our three word classes:
//
int hc1 = hcrf.Compute(hello);
int hc2 = hcrf.Compute(car);
int hc3 = hcrf.Compute(wardrobe);
Assert.AreEqual(0, hc1);
Assert.AreEqual(1, hc2);
Assert.AreEqual(2, hc3);
}
/// <summary>
/// learn Hmm model for samples in given database using Baum Welch unsupervised learning algorithm
/// then used the learned model to classify the training samples
/// </summary>
/// <param name="database"></param>
/// <returns></returns>
static HiddenMarkovClassifier<MultivariateNormalDistribution> learnHMM(Database database)
{
BindingList<Sequence> samples = database.Samples;
BindingList<String> classes = database.Classes;
double[][][] inputs = new double[samples.Count][][];
int[] outputs = new int[samples.Count];
for (int i = 0; i < inputs.Length; i++)
{
inputs[i] = samples[i].Input;
outputs[i] = samples[i].Output;
}
int states = 5;
int iterations = 0;
double tolerance = 0.1;
bool rejection = true;
HiddenMarkovClassifier<MultivariateNormalDistribution> hmm = new HiddenMarkovClassifier<MultivariateNormalDistribution>(classes.Count,
new Forward(states), new MultivariateNormalDistribution(2), classes.ToArray());
// Create the learning algorithm for the ensemble classifier
var teacher = new HiddenMarkovClassifierLearning<MultivariateNormalDistribution>(hmm,
// Train each model using the selected convergence criteria
i => new BaumWelchLearning<MultivariateNormalDistribution>(hmm.Models[i])
{
Tolerance = tolerance,
Iterations = iterations,
FittingOptions = new NormalOptions()
{
Regularization = 1e-5
}
}
);
teacher.Empirical = true;
teacher.Rejection = rejection;
// Run the learning algorithm
double error = teacher.Run(inputs, outputs);
// Classify all training instances
foreach (var sample in database.Samples)
{
sample.RecognizedAs = hmm.Compute(sample.Input);
}
return hmm;
}
public void LearnTest4()
{
// Create a Continuous density Hidden Markov Model Sequence Classifier
// to detect a multivariate sequence and the same sequence backwards.
double[][][] sequences = new double[][][]
{
new double[][]
{
// This is the first sequence with label = 0
new double[] { 0 },
new double[] { 1 },
new double[] { 2 },
new double[] { 3 },
new double[] { 4 },
},
new double[][]
{
// This is the second sequence with label = 1
new double[] { 4 },
new double[] { 3 },
new double[] { 2 },
new double[] { 1 },
new double[] { 0 },
}
};
// Labels for the sequences
int[] labels = { 0, 1 };
// Create a mixture of two 1-dimensional normal distributions (by default,
// initialized with zero mean and unit covariance matrices).
var density = new MultivariateMixture<MultivariateNormalDistribution>(
new MultivariateNormalDistribution(1),
new MultivariateNormalDistribution(1));
// Creates a sequence classifier containing 2 hidden Markov Models with 2 states
// and an underlying multivariate mixture of Normal distributions as density.
var classifier = new HiddenMarkovClassifier<MultivariateMixture<MultivariateNormalDistribution>>(
2, new Ergodic(2), density);
// Configure the learning algorithms to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning<MultivariateMixture<MultivariateNormalDistribution>>(
classifier,
// Train each model until the log-likelihood changes less than 0.0001
modelIndex => new BaumWelchLearning<MultivariateMixture<MultivariateNormalDistribution>>(
classifier.Models[modelIndex])
{
Tolerance = 0.0001,
Iterations = 0,
}
);
// Train the sequence classifier using the algorithm
double logLikelihood = teacher.Run(sequences, labels);
// Calculate the probability that the given
// sequences originated from the model
double likelihood1, likelihood2;
// Try to classify the 1st sequence (output should be 0)
int c1 = classifier.Compute(sequences[0], out likelihood1);
// Try to classify the 2nd sequence (output should be 1)
int c2 = classifier.Compute(sequences[1], out likelihood2);
Assert.AreEqual(0, c1);
Assert.AreEqual(1, c2);
Assert.AreEqual(-13.271981026832933, logLikelihood, 1e-10);
Assert.AreEqual(0.99999791320102149, likelihood1, 1e-10);
Assert.AreEqual(0.99999791320102149, likelihood2, 1e-10);
Assert.IsFalse(double.IsNaN(logLikelihood));
Assert.IsFalse(double.IsNaN(likelihood1));
Assert.IsFalse(double.IsNaN(likelihood2));
}
public int ComputeToInt(double[][] sequence)
{
return(classifer.Compute(sequence));
}
public void LearnTest6()
{
// Create a Continuous density Hidden Markov Model Sequence Classifier
// to detect a multivariate sequence and the same sequence backwards.
double[][][] sequences = new double[][][]
{
new double[][]
{
// This is the first sequence with label = 0
new double[] { 0, 1 },
new double[] { 1, 2 },
new double[] { 2, 3 },
new double[] { 3, 4 },
new double[] { 4, 5 },
},
new double[][]
{
// This is the second sequence with label = 1
new double[] { 4, 3 },
new double[] { 3, 2 },
new double[] { 2, 1 },
new double[] { 1, 0 },
new double[] { 0, -1 },
}
};
// Labels for the sequences
int[] labels = { 0, 1 };
var density = new MultivariateNormalDistribution(2);
// Creates a sequence classifier containing 2 hidden Markov Models with 2 states
// and an underlying multivariate mixture of Normal distributions as density.
var classifier = new HiddenMarkovClassifier<MultivariateNormalDistribution>(
2, new Custom(new double[2, 2], new double[2]), density);
// Configure the learning algorithms to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning<MultivariateNormalDistribution>(
classifier,
// Train each model until the log-likelihood changes less than 0.0001
modelIndex => new BaumWelchLearning<MultivariateNormalDistribution>(
classifier.Models[modelIndex])
{
Tolerance = 0.0001,
Iterations = 0,
FittingOptions = new NormalOptions() { Diagonal = true }
}
);
// Train the sequence classifier using the algorithm
double logLikelihood = teacher.Run(sequences, labels);
// Calculate the probability that the given
// sequences originated from the model
double response1, response2;
// Try to classify the 1st sequence (output should be 0)
int c1 = classifier.Compute(sequences[0], out response1);
// Try to classify the 2nd sequence (output should be 1)
int c2 = classifier.Compute(sequences[1], out response2);
Assert.AreEqual(double.NegativeInfinity, logLikelihood);
Assert.AreEqual(0, response1);
Assert.AreEqual(0, response2);
Assert.IsFalse(double.IsNaN(logLikelihood));
Assert.IsFalse(double.IsNaN(response1));
Assert.IsFalse(double.IsNaN(response2));
}
public Form1()
{
InitializeComponent();
//double[][][] sequences = new double[][][]
//{
// new double[][]
// {
// // This is the first sequence with label = 0
// new double[] { 0, 1 },
// new double[] { 1, 2 },
// new double[] { 2, 3 },
// new double[] { 3, 4 },
// new double[] { 4, 5 },
// },
// new double[][]
// {
// // This is the second sequence with label = 1
// new double[] { 4, 3 },
// new double[] { 3, 2 },
// new double[] { 2, 1 },
// new double[] { 1, 0 },
// new double[] { 0, -1 },
// }
//};
// Labels for the sequences
//int[] labels = { 0, 1 };
#region Old
//Dictionary<string, List<List<double[]>>> dictAll;
//using (System.IO.FileStream fs = new System.IO.FileStream(@".\SkeletonsAsDouble.serialized", System.IO.FileMode.Open))
//{
// BinaryFormatter bf = new BinaryFormatter();
// dictAll = (Dictionary<string, List<List<double[]>>>)bf.Deserialize(fs);
//}
//#region Export to CSV
//using (System.IO.StreamWriter sw = new System.IO.StreamWriter(@".\SkeletonsAsCSV.csv"))
//{
// double[] rgFirstFrame = dictAll[dictAll.Keys.ElementAt(0)][0][0];
// string strHeader = string.Empty;
// for (int i = 0; i < rgFirstFrame.Length; i++)
// {
// strHeader += string.Format("F{0},", i);
// }
// strHeader += "Label";
// sw.WriteLine(strHeader);
// foreach (KeyValuePair<string, List<List<double[]>>> kvp in dictAll)
// {
// foreach (List<double[]> lstGesture in kvp.Value)
// {
// foreach (double[] frame in lstGesture)
// {
// sw.WriteLine(string.Format("{0},{1}", string.Join(",", frame), kvp.Key));
// }
// }
// }
//}
#endregion
#region CHANGE THESE PARAMETERS
const int nStartHiddenCount = 2;
const int nEndHiddenCount = 12;
const int nFeatureCount = 8; // Both DataOrig.dat and Data.dat have dimensionality of 8
#endregion
string strFile;
int nClasses;
#if STRHC1
// Data.dat = Second STRHC (10 classes but not as noisy)
strFile = @".\Data.dat";
nClasses = 10;
#else
// Data1.dat = First STRHC (4 classes but very noisy)
strFile = @".\DataOrig.dat";
nClasses = 4;
#endif
Dictionary <int, Tuple <Dictionary <string, List <List <double[]> > >, Dictionary <string, List <List <double[]> > > > > dict;
using (System.IO.FileStream fs = new System.IO.FileStream(strFile, System.IO.FileMode.Open, System.IO.FileAccess.Read))
{
BinaryFormatter bf = new BinaryFormatter();
dict = bf.Deserialize(fs) as Dictionary <int, Tuple <Dictionary <string, List <List <double[]> > >, Dictionary <string, List <List <double[]> > > > >;
}
using (System.IO.StreamWriter sw = new System.IO.StreamWriter(@"HMMOutput.txt"))
{
for (int nHiddenCount = nStartHiddenCount; nHiddenCount <= nEndHiddenCount; nHiddenCount++)
{
double fAverage = 0.0;
int nEpoch = 0;
List <int> lstTrainTimes = new List <int>();
List <int> lstRecogTimes = new List <int>();
//for (int i = 1; i <= nEpochs; i++)
foreach (KeyValuePair <int, Tuple <Dictionary <string, List <List <double[]> > >, Dictionary <string, List <List <double[]> > > > > kvpTT in dict)
{
nEpoch++;
#region Old
//Dictionary<string, List<List<double[]>>> dictTrain = new Dictionary<string, List<List<double[]>>>();
//Dictionary<string, List<List<double[]>>> dictTest = new Dictionary<string, List<List<double[]>>>();
//#region Divide the vectors into training and testing sets
//foreach (KeyValuePair<string, List<List<double[]>>> kvp in dictAll)
//{
// int nGroupSize = kvp.Value.Count / nEpochs;
// List<List<double[]>> lstTrain;
// List<List<double[]>> lstTest;
// dictTrain.Add(kvp.Key, (lstTrain = new List<List<double[]>>()));
// dictTest.Add(kvp.Key, (lstTest = new List<List<double[]>>()));
// for (int j = 0; j < kvp.Value.Count; j++)
// {
// if (j < i * nGroupSize && j >= (i * nGroupSize) - nGroupSize)
// {
// lstTest.Add(kvp.Value[j]);
// }
// else
// {
// lstTrain.Add(kvp.Value[j]);
// }
// }
//}
//#endregion
#endregion
Dictionary <string, List <List <double[]> > > dictTrain = kvpTT.Value.Item1;
Dictionary <string, List <List <double[]> > > dictTest = kvpTT.Value.Item2;
double[][][] sequences = new double[dictTrain.Sum(kvp => kvp.Value.Count)][][];
int[] labels = new int[sequences.Length];
#region The Sequences
int nIndex = 0;
foreach (KeyValuePair <string, List <List <double[]> > > kvp in dictTrain)
{
foreach (List <double[]> lst in kvp.Value)
{
sequences[nIndex] = lst.ToArray();
labels[nIndex] = Array.IndexOf(dictTrain.Keys.ToArray(), kvp.Key);
nIndex++;
}
}
#endregion
var initialDensity = new MultivariateNormalDistribution(nFeatureCount);
// Creates a sequence classifier containing 2 hidden Markov Models with 2 states
// and an underlying multivariate mixture of Normal distributions as density.
var classifier = new HiddenMarkovClassifier <MultivariateNormalDistribution>(
classes: nClasses, topology: new Forward(nHiddenCount), initial: initialDensity);
// Configure the learning algorithms to train the sequence classifier
var teacher = new HiddenMarkovClassifierLearning <MultivariateNormalDistribution>(
classifier,
// Train each model until the log-likelihood changes less than 0.0001
modelIndex => new BaumWelchLearning <MultivariateNormalDistribution>(
classifier.Models[modelIndex])
{
Tolerance = 0.0001,
Iterations = 0,
FittingOptions = new NormalOptions()
{
Diagonal = true, // only diagonal covariance matrices
Regularization = 1e-5 // avoid non-positive definite errors
}
}
);
// Train the sequence classifier using the algorithm
System.Diagnostics.Stopwatch watchTrain = new System.Diagnostics.Stopwatch();
watchTrain.Start();
double logLikelihood = teacher.Run(sequences, labels);
watchTrain.Stop();
lstTrainTimes.Add((int)watchTrain.ElapsedMilliseconds);
//// Calculate the probability that the given
//// sequences originated from the model
//double likelihood, likelihood2;
//// Try to classify the 1st sequence (output should be 0)
//int c1 = classifier.Compute(sequences[0], out likelihood);
//// Try to classify the 2nd sequence (output should be 1)
//int c2 = classifier.Compute(sequences[1], out likelihood2);
sw.WriteLine("Epoch: {0} -- Hidden State: {1}\t\tTotal Train Time: {2}", nEpoch, nHiddenCount, watchTrain.ElapsedMilliseconds.ToString());
int nCorrect = 0;
int nIncorrect = 0;
foreach (KeyValuePair <string, List <List <double[]> > > kvp in dictTest)
{
foreach (List <double[]> lst in kvp.Value)
{
System.Diagnostics.Stopwatch watch = new System.Diagnostics.Stopwatch();
watch.Start();
int nClassIndex = classifier.Compute(lst.ToArray());
watch.Stop();
sw.WriteLine(string.Format("Should be: {0}\tRecognized: {1}\t\tTime: {2} ms", kvp.Key, dictTest.Keys.ElementAt(nClassIndex), watch.ElapsedMilliseconds));
lstRecogTimes.Add((int)watch.ElapsedMilliseconds);
if (dictTest.Keys.ElementAt(nClassIndex) == kvp.Key)
{
nCorrect++;
}
else
{
nIncorrect++;
}
}
}
fAverage += (double)nCorrect / (nCorrect + nIncorrect);
sw.WriteLine(string.Format("Correct: {0} of {1} ({2:P3})", nCorrect, nCorrect + nIncorrect, (double)nCorrect / (nCorrect + nIncorrect)));
sw.WriteLine();
sw.WriteLine();
}
sw.WriteLine("Average Correct for {0} Hidden: {1:P3}", nHiddenCount, fAverage / 5);
sw.WriteLine("Average Train Time for {0}: {1:F2}", nHiddenCount, lstTrainTimes.Select(v => (double)v).Average());
sw.WriteLine("Std. Dev. Train Time for {0}: {1:F2}", nHiddenCount, lstTrainTimes.Select(v => (double)v).StandardDeviation());
sw.WriteLine("Average Recog. Time for {0}: {1:F2}", nHiddenCount, lstRecogTimes.Select(v => (double)v).Average());
sw.WriteLine("Std. Dev. Recog. Time for {0}: {1:F2}", nHiddenCount, lstRecogTimes.Select(v => (double)v).StandardDeviation());
sw.WriteLine();
sw.WriteLine();
sw.WriteLine();
sw.WriteLine();
}
}
}
