public void AddInboundConnection_ValidatesArgs()
{
// Setup
var neuron = new BackpropagationNeuron(0.0d, new Mock<IActivationFunction>().Object);
// Execute/Verify
neuron.AddInboundConnection(null);
}
public void ApplyWeightAdjustments_PropogatesThroughNetwork()
{
//// SETUP
const double CachedOutput = 1.23d;
const double ErrorSignal = -2.3d;
const double Derivative = 4.567d;
const float Momentum = 0.9f;
const float LearningRate = 0.1f;
// Create 2 inbound and 2 outbound mock connections.
var mockInbound1 = new Mock<ISupervisedLearnerConnection>();
var mockInbound2 = new Mock<ISupervisedLearnerConnection>();
var mockOutbound1 = new Mock<ISupervisedLearnerConnection>();
var mockOutbound2 = new Mock<ISupervisedLearnerConnection>();
// program the mock outbounds such that both are reporting error signals
mockOutbound1.SetupGet(mock => mock.IsReportingError).Returns(true);
mockOutbound2.SetupGet(mock => mock.IsReportingError).Returns(true);
// mock activation function
var mockActivationFunction = new Mock<IActivationFunction>();
mockActivationFunction.Setup(mock => mock.Invoke(It.IsAny<double>())).Returns(CachedOutput);
mockActivationFunction.Setup(mock => mock.InvokeDerivative(CachedOutput)).Returns(Derivative);
// Create the test object.
var neuron = new BackpropagationNeuron(0.0d, mockActivationFunction.Object);
neuron.AddInboundConnection(mockInbound1.Object);
neuron.AddInboundConnection(mockInbound2.Object);
neuron.AddOutboundConnection(mockOutbound1.Object);
neuron.AddOutboundConnection(mockOutbound2.Object);
neuron.Fire(new[] { 0.0d });
neuron.CalculateError(ErrorSignal);
// EXECUTION
neuron.ApplyWeightAdjustments(LearningRate, Momentum);
// VERIFICATION
mockOutbound1.Verify(mock => mock.ApplyWeightAdjustments(LearningRate, Momentum), Times.Once());
mockOutbound2.Verify(mock => mock.ApplyWeightAdjustments(LearningRate, Momentum), Times.Once());
}
public void ClearCachedErrors_ClearsCachedError()
{
//// SETUP
const double CachedOutput = 1.23d;
const double ErrorSignal = -2.3d;
const double Derivative = 4.567d;
// Create 2 inbound and 2 outbound mock connections.
var mockInbound1 = new Mock<ISupervisedLearnerConnection>();
var mockInbound2 = new Mock<ISupervisedLearnerConnection>();
var mockOutbound1 = new Mock<ISupervisedLearnerConnection>();
var mockOutbound2 = new Mock<ISupervisedLearnerConnection>();
// program the mock outbounds such that both are reporting error signals
mockOutbound1.SetupGet(mock => mock.IsReportingError).Returns(true);
mockOutbound2.SetupGet(mock => mock.IsReportingError).Returns(true);
// mock activation function
var mockActivationFunction = new Mock<IActivationFunction>();
mockActivationFunction.Setup(mock => mock.Invoke(It.IsAny<double>())).Returns(CachedOutput);
mockActivationFunction.Setup(mock => mock.InvokeDerivative(CachedOutput)).Returns(Derivative);
// Create the test object.
var neuron = new BackpropagationNeuron(0.0d, mockActivationFunction.Object);
neuron.AddInboundConnection(mockInbound1.Object);
neuron.AddInboundConnection(mockInbound2.Object);
neuron.AddOutboundConnection(mockOutbound1.Object);
neuron.AddOutboundConnection(mockOutbound2.Object);
neuron.Fire(new[] { 0.0d });
// EXECUTION
neuron.CalculateError(ErrorSignal);
var cachedError = neuron.CachedErrors[0];
neuron.ClearCachedErrors();
var clearedError = neuron.CachedErrors[0];
// VERIFICATION
Assert.AreEqual(1, neuron.CachedErrors.Length);
Assert.AreNotEqual(0.0d, cachedError);
Assert.AreEqual(0.0d, clearedError, Epsilon);
}
public void CalculateError_DoesNotCalculateUntilAllConnectionsReport()
{
//// SETUP
// Create 2 inbound and 2 outbound mock connections.
var mockInbound1 = new Mock<ISupervisedLearnerConnection>();
var mockInbound2 = new Mock<ISupervisedLearnerConnection>();
var mockOutboundReporting = new Mock<ISupervisedLearnerConnection>();
var mockOutboundNotReporting = new Mock<ISupervisedLearnerConnection>();
// program the mock outbounds such that one is reporting and one isn't
mockOutboundReporting.SetupGet(mock => mock.IsReportingError).Returns(true);
mockOutboundNotReporting.SetupGet(mock => mock.IsReportingError).Returns(false);
// mock activation function
var mockActivationFunction = new Mock<IActivationFunction>();
// Create the test object.
var neuron = new BackpropagationNeuron(0.0d, mockActivationFunction.Object);
neuron.AddInboundConnection(mockInbound1.Object);
neuron.AddInboundConnection(mockInbound2.Object);
neuron.AddOutboundConnection(mockOutboundReporting.Object);
neuron.AddOutboundConnection(mockOutboundNotReporting.Object);
// EXECUTION
const double ErrorSignal = -2.3d;
neuron.CalculateError(ErrorSignal);
// VERIFICATION: The IsReporting signals were checked...
mockOutboundReporting.Verify(mock => mock.IsReportingError, Times.Exactly(1));
mockOutboundNotReporting.Verify(mock => mock.IsReportingError, Times.Exactly(1));
// ...but no calculation activities occurred.
mockActivationFunction.Verify(mock => mock.InvokeDerivative(It.IsAny<double>()), Times.Never());
mockInbound1.Verify(mock => mock.ReportError(It.IsAny<double>()), Times.Never());
mockInbound2.Verify(mock => mock.ReportError(It.IsAny<double>()), Times.Never());
mockOutboundReporting.Verify(mock => mock.ClearReportingFlag(), Times.Never());
mockOutboundNotReporting.Verify(mock => mock.ClearReportingFlag(), Times.Never());
Assert.AreEqual(1, neuron.CachedErrors.Length);
Assert.AreEqual(ErrorSignal, neuron.CachedErrors[0], Epsilon);
}
public void CalculateError_CalculatesAfterAllConnectionsHaveReported()
{
//// SETUP
const double CachedOutput = 1.23d;
const double ErrorSignal1 = -2.3d;
const double ErrorSignal2 = -9.87d;
const double AccumulatedError = ErrorSignal1 + ErrorSignal2;
const double Derivative = 4.567d;
const double ErrorDelta = AccumulatedError * Derivative;
// Create 2 inbound and 2 outbound mock connections.
var mockInbound1 = new Mock<ISupervisedLearnerConnection>();
var mockInbound2 = new Mock<ISupervisedLearnerConnection>();
var mockOutbound1 = new Mock<ISupervisedLearnerConnection>();
var mockOutbound2 = new Mock<ISupervisedLearnerConnection>();
// program the mock outbounds such that one is reporting and one isn't
mockOutbound1.SetupGet(mock => mock.IsReportingError).Returns(true);
mockOutbound2.SetupGet(mock => mock.IsReportingError).Returns(false);
// mock activation function
var mockActivationFunction = new Mock<IActivationFunction>();
mockActivationFunction.Setup(mock => mock.Invoke(It.IsAny<double>())).Returns(CachedOutput);
mockActivationFunction.Setup(mock => mock.InvokeDerivative(CachedOutput)).Returns(Derivative);
// Create the test object.
var neuron = new BackpropagationNeuron(0.0d, mockActivationFunction.Object);
neuron.AddInboundConnection(mockInbound1.Object);
neuron.AddInboundConnection(mockInbound2.Object);
neuron.AddOutboundConnection(mockOutbound1.Object);
neuron.AddOutboundConnection(mockOutbound2.Object);
// EXECUTION
neuron.Fire(new[] { 0.0d });
neuron.CalculateError(ErrorSignal1);
mockOutbound2.SetupGet(mock => mock.IsReportingError).Returns(true);
neuron.CalculateError(ErrorSignal2);
// VERIFICATION: The IsReporting signals were checked...
mockOutbound1.Verify(mock => mock.IsReportingError, Times.Exactly(2));
mockOutbound2.Verify(mock => mock.IsReportingError, Times.Exactly(2));
// ...and calculation activities occurred.
mockActivationFunction.Verify(mock => mock.InvokeDerivative(CachedOutput), Times.Once());
mockInbound1.Verify(mock => mock.ReportError(ErrorDelta), Times.Once());
mockInbound2.Verify(mock => mock.ReportError(ErrorDelta), Times.Once());
mockOutbound1.Verify(mock => mock.ClearReportingFlag(), Times.Once());
mockOutbound2.Verify(mock => mock.ClearReportingFlag(), Times.Once());
Assert.AreEqual(1, neuron.CachedErrors.Length);
Assert.AreEqual(AccumulatedError, neuron.CachedErrors[0], Epsilon);
}
public void ApplyWeightAdjustments_UpdatesBias()
{
//// SETUP
const double CachedOutput = 1.23d;
const double ErrorSignal1A = -2.3d;
const double ErrorSignal1B = -9.87d;
const double AccumulatedError1 = ErrorSignal1A + ErrorSignal1B;
const double ErrorSignal2 = 3.2345d;
const double Derivative = 4.567d;
const double ErrorDelta1 = AccumulatedError1 * Derivative;
const double ErrorDelta2 = ErrorSignal2 * Derivative;
const float Momentum = 0.9f;
const float LearningRate = 0.1f;
const double Expected1 = ErrorDelta1 * LearningRate;
const double Expected2 = Expected1 + (ErrorDelta2 * LearningRate) + (Momentum * Expected1);
// Create 2 inbound and 2 outbound mock connections.
var mockInbound1 = new Mock<ISupervisedLearnerConnection>();
var mockInbound2 = new Mock<ISupervisedLearnerConnection>();
var mockOutbound1 = new Mock<ISupervisedLearnerConnection>();
var mockOutbound2 = new Mock<ISupervisedLearnerConnection>();
// program the mock outbounds such that one is reporting and one isn't
mockOutbound1.SetupGet(mock => mock.IsReportingError).Returns(true);
mockOutbound2.SetupGet(mock => mock.IsReportingError).Returns(false);
// mock activation function
var mockActivationFunction = new Mock<IActivationFunction>();
mockActivationFunction.Setup(mock => mock.Invoke(It.IsAny<double>())).Returns(CachedOutput);
mockActivationFunction.Setup(mock => mock.InvokeDerivative(CachedOutput)).Returns(Derivative);
// Create the test object.
var neuron = new BackpropagationNeuron(0.0d, mockActivationFunction.Object);
neuron.AddInboundConnection(mockInbound1.Object);
neuron.AddInboundConnection(mockInbound2.Object);
neuron.AddOutboundConnection(mockOutbound1.Object);
neuron.AddOutboundConnection(mockOutbound2.Object);
neuron.Fire(new[] { 0.0d });
// EXECUTION
neuron.CalculateError(ErrorSignal1A);
mockOutbound2.SetupGet(mock => mock.IsReportingError).Returns(true);
neuron.CalculateError(ErrorSignal1B);
neuron.ApplyWeightAdjustments(LearningRate, Momentum);
var actual1 = neuron.Bias;
neuron.ClearCachedErrors();
neuron.CalculateError(ErrorSignal2);
neuron.ApplyWeightAdjustments(LearningRate, Momentum);
var actual2 = neuron.Bias;
// VERIFICATION
Assert.AreEqual(Expected1, actual1, Epsilon);
Assert.AreEqual(Expected2, actual2, Epsilon);
}