public void InitializeTheta()
{
var engine = new NeuralNetworkEngine(_config400x25x10);
var theta = engine.InitializeTheta();
theta.Count.Should().Be(25*401 + 10*26);
theta.All(i => i > 0 && i < 1);
}
public NeuralNetwork(NeuralNetworkConfig config)
{
if (config == null)
{
throw new ArgumentNullException("config");
}
_config = config;
_engine = new NeuralNetworkEngine(config);
}
public void Sigmoid()
{
var engine = new NeuralNetworkEngine(_config400x25x10);
var sigmoid = engine.Sigmoid(new DenseVector(new double[] { 100, 0, -100 }));
sigmoid.Should().NotBeNull().
And.HaveCount(3);
sigmoid[0].Should().BeApproximately(1, 0.01);
sigmoid[1].Should().Be(0.5);
sigmoid[2].Should().BeApproximately(0, 0.01);
}
public void ReshapeTheta()
{
var engine = new NeuralNetworkEngine(_config1x2X2x1);
var vector = new double[] { 11, 12, 21, 22, 11, 12, 13, 11, 12, 21, 22, 31, 32 };
var matrices = engine.ReshapeTheta(vector.ToArray(), new int[] { 1, 2, 1, 3 });
var reshaped = engine.ReshapeTheta(matrices);
reshaped.Should().BeEquivalentTo(vector.ToArray());
var theta = engine.InitializeTheta();
var Theta = engine.ReshapeTheta(theta.ToArray(), _config1x2X2x1.Layers);
Theta.Length.Should().Be(3);
Theta[0].RowCount.Should().Be(2);
Theta[0].ColumnCount.Should().Be(2);
Theta[1].RowCount.Should().Be(2);
Theta[1].ColumnCount.Should().Be(3);
Theta[2].RowCount.Should().Be(1);
Theta[2].ColumnCount.Should().Be(3);
var thetaVector = engine.ReshapeTheta(Theta);
thetaVector.Should().BeEquivalentTo(theta.ToArray());
}
public void SigmoidGradient()
{
var engine = new NeuralNetworkEngine(_config400x25x10);
var sigmoid = engine.SigmoidGradient(new DenseVector(new double[] { 0 }));
sigmoid.Should().NotBeNull().
And.HaveCount(1);
sigmoid[0].Should().Be(0.25);
}
public NeuralNetwork(NeuralNetworkConfig config)
{
if (config == null) throw new ArgumentNullException("config");
_config = config;
_engine = new NeuralNetworkEngine(config);
}
