public void GivenAFourByFourInputAndATwoByTwoFilter_WhenTheNeuronIsActivated_ItShouldReturnAFeatureMap()
{
FeatureMapNeuronActivationStrategy strategy = new FeatureMapNeuronActivationStrategy(MockDotProduct(), new Configuration());
double[,] matrix = new double[3, 3];
matrix[0, 0] = 0;
matrix[0, 1] = 1;
matrix[0, 2] = 2;
matrix[1, 0] = 3;
matrix[1, 1] = 4;
matrix[1, 2] = 5;
matrix[2, 0] = 6;
matrix[2, 1] = 7;
matrix[2, 2] = 8;
ImageInput input = new ImageInput(new TwoDimensionalMapper(), new Configuration());
input.SetNumericRepresentation(matrix);
double[] mockFilter = new double[4];
double[] featureMap = strategy.Activate(input, mockFilter);
double[] expected = { 0.3, 2.7, 6.2, 1.9 };
Assert.IsTrue(Enumerable.SequenceEqual(expected, featureMap));
}
public void GivenAnInputSubsectionAndAFilter_WhenAFeatureMapPartIsCalculated_ItShouldCallTheDotProductConvoluteFunction()
{
double[,] inputSubsection = new double[3, 3];
inputSubsection[0, 0] = 0;
inputSubsection[0, 1] = 1;
inputSubsection[1, 0] = 2;
inputSubsection[1, 1] = 4;
double[] filter = { 0.1, 0.2 };
ImageInput input = new ImageInput(new TwoDimensionalMapper(), new Configuration());
input.SetNumericRepresentation(inputSubsection);
Mock <IDotProductConvolutionStrategy> dotProductMock = new Mock <IDotProductConvolutionStrategy>();
dotProductMock.SetupSequence(x => x.Calculate(It.IsAny <double[]>(), It.IsAny <double[]>()))
.Returns(0.3).Returns(2.7)
.Returns(6.2).Returns(1.9)
.Returns(3.4).Returns(0.8);
FeatureMapNeuronActivationStrategy strategy = new FeatureMapNeuronActivationStrategy(dotProductMock.Object, new Configuration());
double part = strategy.ConvolveSubSection(input.GetNumericRepresentation(), filter);
dotProductMock.Verify(c => c.Calculate(It.IsAny <double[]>(), It.IsAny <double[]>()), Times.Exactly(1));
}
public void GivenAnInputSubsectionAndAFilter_WhenAFeatureMapPartIsCalculated_ItShouldReturnTheDotProduct()
{
double[,] inputSubsection = new double[3, 3];
inputSubsection[0, 0] = 0;
inputSubsection[0, 1] = 1;
inputSubsection[1, 0] = 2;
inputSubsection[1, 1] = 4;
double[] filter = { 0.1, 0.2 };
ImageInput input = new ImageInput(new TwoDimensionalMapper(), new Configuration());
input.SetNumericRepresentation(inputSubsection);
FeatureMapNeuronActivationStrategy strategy = new FeatureMapNeuronActivationStrategy(MockDotProduct(), new Configuration());
double part = strategy.ConvolveSubSection(input.GetNumericRepresentation(), filter);
Assert.AreEqual(0.3, part);
}
