public void NeuralNetEvaluation()
{
// Arrange
var network = new NeuralNet(2, 1);
var hidden = new NeuralNetLayer(2);
network.AddHiddenLayer(hidden);
var weights = new List <decimal> {
1, -1, 0.5M, -0.5M, 1, -1
};
var values = new List <decimal> {
1, -1
};
// Act
network.SetWeights(weights);
network.InputLayer.SetValues(values);
network.Evaluate();
// Assert
network.InputLayer.Neurons.First().Value.ShouldBe(1);
network.InputLayer.Neurons.Last().Value.ShouldBe(-1);
hidden.Neurons.First().Value.ShouldBe(1);
hidden.Neurons.Last().Value.ShouldBe(0.5M);
network.OutputLayer.Neurons.Single().Value.ShouldBe(0.25M);
}
public void NeuralNetLayerShouldHaveCorrectNumberOfNodes(int expectedNodes)
{
// Arrange / Act
var layer = new NeuralNetLayer(expectedNodes);
// Assert
layer.Neurons.ShouldNotBeNull();
layer.Neurons.Count().ShouldBe(expectedNodes);
layer.Neurons.Each(n => n.ShouldNotBeNull());
}
public void HiddenLayersShouldBeConnectionTargets()
{
// Arrange
var numInputLayer = 3;
var network = new NeuralNet(3, 1);
var hidden = new NeuralNetLayer(2);
// Act
network.AddHiddenLayer(hidden);
network.Connect();
// Assert
hidden.Neurons.Each(n => n.Inputs.Count().ShouldBe(numInputLayer));
}
public void SetNodeValuesShouldModifyNodes()
{
// Arrange
var layer = new NeuralNetLayer(3);
var values = new List <decimal> {
1, 2, 3
};
// Act
layer.SetValues(values);
// Assert
var neurons = layer.Neurons.ToList();
int index = 0;
values.Each(v => neurons[index++].Value.ShouldBe(v));
}
public void NodeEvaluationShouldSetNewNodeValue()
{
// Arrange
var input = new NeuralNetLayer(1);
var target = new NeuralNetLayer(1);
input.Neurons.First().Connect(target.Neurons.First());
input.Neurons.First().Value = 0.5M;
// Act
var evalResult = target.Evaluate();
// Assert
target.Neurons.First().Inputs.Single().Weight.ShouldBe(1M);
target.Neurons.First().Inputs.Single().Source.ShouldBe(input.Neurons.First());
target.Neurons.First().Value.ShouldBe(0.5M);
}
public void HiddenLayersShouldBeIncluded()
{
// Arrange
var network = new NeuralNet(3, 1);
var hidden = new NeuralNetLayer(5);
// Act
network.AddHiddenLayer(hidden
);
// Assert
var layers = network.Layers.ToList();
layers.Count.ShouldBe(3);
layers[0].ShouldBe(network.InputLayer);
layers[1].ShouldBe(hidden);
layers[2].ShouldBe(network.OutputLayer);
}
public void SetWeightsShouldWork()
{
// Arrange
var network = new NeuralNet(2, 1);
var hidden = new NeuralNetLayer(2);
network.AddHiddenLayer(hidden);
var weights = new List <decimal> {
1, -1, 0.5M, -0.5M, 1, -1
};
// Act
network.SetWeights(weights);
// Assert
hidden.Neurons.First().Inputs.First().Weight.ShouldBe(1);
hidden.Neurons.First().Inputs.Last().Weight.ShouldBe(-1);
hidden.Neurons.Last().Inputs.First().Weight.ShouldBe(0.5M);
hidden.Neurons.Last().Inputs.Last().Weight.ShouldBe(-0.5M);
network.OutputLayer.Neurons.Single().Inputs.First().Weight.ShouldBe(1);
network.OutputLayer.Neurons.Single().Inputs.Last().Weight.ShouldBe(-1);
}