public void PredictWinner_NotTrained_ReturnUnknownClass()
{
// Given
var testPattern = new SparsePattern(new Dictionary<int, float> { { 1, 1.0f }, { 2, 0.85f } }, 1);
// When
var winningClass = m_classifier.PredictWinner(testPattern);
// Then
Assert.That(winningClass, Is.EqualTo(UnknownClassID));
}
public void PredictWinner_ClassTwoWins()
{
// Given
const int classOne = 1;
const int classTwo = 2;
var patterns = new[]
{
new SparsePattern(new Dictionary<int, float> {{1,1.0f}, {2,1.0f} }, classOne),
new SparsePattern(new Dictionary<int, float> {{3,1.0f}, {4,1.0f}}, classTwo)
};
var testPattern = new SparsePattern(new Dictionary<int, float> { {3, 2.0f }, {4, 4.0f } }, classTwo);
// When
m_classifier.Train(patterns);
var winningClass = m_classifier.PredictWinner(testPattern);
// Then
Assert.That(winningClass, Is.EqualTo(classTwo));
}
public void Write(SparsePattern sparsePattern)
{
if(sparsePattern.Explain == null) return;
if (!HeaderWritten)
{
foreach (var explanation in sparsePattern.Explain)
{
m_streamWriter.WriteLine("classes({0}).name = '{1}';", explanation.Key, explanation.Value);
}
HeaderWritten = true;
}
var features = new float[m_maxFeatureId + 1];
features[0] = sparsePattern.Label;
for (var featId = 0; featId < m_maxFeatureId; featId++)
{
features[featId + 1] = sparsePattern.Features[featId + 1];
}
var featuresAsString = (from f in features select f.ToString(m_formatProvider)).ToArray();
m_streamWriter.WriteLine("pattern({0},:) = [{1}];", ++m_patternId, string.Join(",", featuresAsString));
}
