public void CloneTest()
{
ReLULayer layer1 = new ReLULayer(new Shape(new int[] { 2 }));
ReLULayer layer2 = layer1.Clone() as ReLULayer;
Assert.AreEqual(JsonConvert.SerializeObject(layer1), JsonConvert.SerializeObject(layer2));
}
public void ForwardBackwardTest()
{
Shape shape = new Shape(new int[] { 2 });
ReLULayer layer = new ReLULayer(shape);
Session session = new Session();
Tensor source = new Tensor(null, shape);
source.Set(new float[] { 2, -3 });
Tensor x = source.Clone() as Tensor;
Tensor y = layer.Forward(session, new[] { x })[0];
float[] expected = source.Weights.Take(source.Length).Select(w => ReLULayerTest.activation(w)).ToArray();
Helpers.AreArraysEqual(x.Length, expected, y.Weights);
// unroll the graph
float[] dy = Enumerable.Range(1, x.Length).Select(w => (float)w).ToArray();
y.SetGradient(dy);
session.Unroll();
Helpers.AreArraysEqual(
expected.Length,
expected.Zip(dy, (w, dw) => ReLULayerTest.derivative(w) * dw).ToArray(),
x.Gradient);
}
public void CopyConstructorTest1()
{
ReLULayer layer1 = new ReLULayer(new Shape(new int[] { 2 }));
ReLULayer layer2 = new ReLULayer(layer1);
Assert.AreEqual(JsonConvert.SerializeObject(layer1), JsonConvert.SerializeObject(layer2));
}
public void ArchitectureConstructorTest1()
{
Shape shape = new Shape(new int[] { 2 });
ReLULayer layer = new ReLULayer(shape, "RELU", null);
CollectionAssert.AreEqual(shape.Axes, layer.OutputShape.Axes);
Assert.AreEqual("RELU", layer.Architecture);
}
public void SerializeTest()
{
ReLULayer layer1 = new ReLULayer(new Shape(new int[] { 2 }));
string s1 = JsonConvert.SerializeObject(layer1);
ReLULayer layer2 = JsonConvert.DeserializeObject <ReLULayer>(s1);
string s2 = JsonConvert.SerializeObject(layer2);
Assert.AreEqual(s1, s2);
}
public MNISTNetwork(int inputSize, int hiddenSize, int outputSize) : base()
{
this.inputSize = inputSize;
this.hiddenSize = hiddenSize;
this.outputSize = outputSize;
affine1 = new AffineLayer(inputSize, hiddenSize);
relu = new ReLULayer();
affine2 = new AffineLayer(hiddenSize, outputSize);
softmax = new SoftmaxLayer();
optimizer = new MomentumOptimizer(0.9f);
}
public void ArchitectureConstructorTest2()
{
string architecture = "REL";
try
{
ReLULayer layer = new ReLULayer(new Shape(new int[] { 2 }), architecture, null);
}
catch (ArgumentException e)
{
Assert.AreEqual(
new ArgumentException(string.Format(CultureInfo.InvariantCulture, Properties.Resources.E_InvalidLayerArchitecture, architecture), nameof(architecture)).Message,
e.Message);
throw;
}
}
public MNISTBatchNormalizationNetwork(int inputSize, int hiddenSize, int outputSize) : base()
{
this.inputSize = inputSize;
this.hiddenSize = hiddenSize;
this.outputSize = outputSize;
affine1 = new AffineLayer(inputSize, hiddenSize, Mathf.Sqrt(2.0f / inputSize));
bn1 = new BatchNormalizationLayer(hiddenSize, hiddenSize);
relu1 = new ReLULayer();
affine2 = new AffineLayer(hiddenSize, hiddenSize, Mathf.Sqrt(2.0f / hiddenSize));
bn2 = new BatchNormalizationLayer(hiddenSize, hiddenSize);
relu2 = new ReLULayer();
affine3 = new AffineLayer(hiddenSize, outputSize, Mathf.Sqrt(2.0f / hiddenSize));
softmax = new SoftmaxLayer();
optimizer = new MomentumOptimizer(0.9f);
}
