public void CloneTest()
{
LRNLayer layer1 = new LRNLayer(LRNLayerTest.sourceShape, 7);
LRNLayer layer2 = layer1.Clone() as LRNLayer;
Assert.AreEqual(JsonConvert.SerializeObject(layer1), JsonConvert.SerializeObject(layer2));
}
public void CopyConstructorTest1()
{
LRNLayer layer1 = new LRNLayer(LRNLayerTest.sourceShape, 7);
LRNLayer layer2 = new LRNLayer(layer1);
Assert.AreEqual(JsonConvert.SerializeObject(layer1), JsonConvert.SerializeObject(layer2));
}
public void SerializeTest()
{
LRNLayer layer1 = new LRNLayer(LRNLayerTest.sourceShape, 7);
string s1 = JsonConvert.SerializeObject(layer1);
LRNLayer layer2 = JsonConvert.DeserializeObject <LRNLayer>(s1);
string s2 = JsonConvert.SerializeObject(layer2);
Assert.AreEqual(s1, s2);
}
public void ConstructorTest2()
{
LRNLayer layer = new LRNLayer(LRNLayerTest.sourceShape, 7);
Assert.AreEqual(1, layer.NumberOfOutputs);
CollectionAssert.AreEqual(LRNLayerTest.sourceShape.Axes, layer.OutputShape.Axes);
Assert.AreEqual("LRN7", layer.Architecture);
Assert.AreEqual(7, layer.KernelSize);
Assert.AreEqual(LRNLayer.DefaultAlpha, layer.Alpha);
Assert.AreEqual(LRNLayer.DefaultBeta, layer.Beta);
Assert.AreEqual(LRNLayer.DefaultK, layer.K);
}
public void ConstructorTest1()
{
LRNLayer layer = new LRNLayer(LRNLayerTest.sourceShape, 7, 0.001f, 0.5f, 3f);
Assert.AreEqual(1, layer.NumberOfOutputs);
CollectionAssert.AreEqual(LRNLayerTest.sourceShape.Axes, layer.OutputShape.Axes);
Assert.AreEqual("LRN7(A=0.001;B=0.5;K=3)", layer.Architecture);
Assert.AreEqual(7, layer.KernelSize);
Assert.AreEqual(0.001f, layer.Alpha);
Assert.AreEqual(0.5f, layer.Beta);
Assert.AreEqual(3f, layer.K);
}
public void ArchitectureConstructorTest1()
{
const string Architecture = "LRN5";
LRNLayer layer = new LRNLayer(LRNLayerTest.sourceShape, Architecture, null);
Assert.AreEqual(1, layer.NumberOfOutputs);
CollectionAssert.AreEqual(LRNLayerTest.sourceShape.Axes, layer.OutputShape.Axes);
Assert.AreEqual(Architecture, layer.Architecture);
Assert.AreEqual(5, layer.KernelSize);
Assert.AreEqual(LRNLayer.DefaultAlpha, layer.Alpha);
Assert.AreEqual(LRNLayer.DefaultBeta, layer.Beta);
Assert.AreEqual(LRNLayer.DefaultK, layer.K);
}
public void ArchitectureConstructorTest2()
{
const string Architecture = "LRN7(A=0.001;B=0.5;K=3)";
LRNLayer layer = new LRNLayer(LRNLayerTest.sourceShape, Architecture, null);
Assert.AreEqual(1, layer.NumberOfOutputs);
CollectionAssert.AreEqual(LRNLayerTest.sourceShape.Axes, layer.OutputShape.Axes);
Assert.AreEqual(Architecture, layer.Architecture);
Assert.AreEqual(7, layer.KernelSize);
Assert.AreEqual(0.001f, layer.Alpha);
Assert.AreEqual(0.5f, layer.Beta);
Assert.AreEqual(3f, layer.K);
}
public void ArchitectureConstructorTest3()
{
string architecture = "LRN";
try
{
LRNLayer layer = new LRNLayer(LRNLayerTest.sourceShape, architecture, null);
}
catch (ArgumentException e)
{
Assert.AreEqual(
new ArgumentException(string.Format(CultureInfo.InvariantCulture, Properties.Resources.E_InvalidLayerArchitecture, architecture), nameof(architecture)).Message,
e.Message);
throw;
}
}
private static float Forward(LRNLayer layer, params float[] xs)
{
float scale = layer.K + (layer.Alpha * xs.Sum(x => x * x) / layer.KernelSize);
return(xs[0] * (float)Math.Pow(scale, -layer.Beta));
}
public void ForwardBackwardTest()
{
Shape shape = new Shape(Shape.BWHC, -1, 1, 2, 5);
LRNLayer layer = new LRNLayer(shape, 3);
Tensor xTemp = new Tensor(null, shape.Reshape(Axis.B, 1));
xTemp.Set(new float[]
{
1, 11, 21, 31, 41, 2, 12, 22, 32, 42
});
Tensor expectedTemp = new Tensor(null, layer.OutputShape.Reshape(Axis.B, 1));
expectedTemp.Set(new float[]
{
LRNLayerTest.Forward(layer, 1, 11),
LRNLayerTest.Forward(layer, 11, 1, 21),
LRNLayerTest.Forward(layer, 21, 11, 31),
LRNLayerTest.Forward(layer, 31, 21, 41),
LRNLayerTest.Forward(layer, 41, 31),
LRNLayerTest.Forward(layer, 2, 12),
LRNLayerTest.Forward(layer, 12, 2, 22),
LRNLayerTest.Forward(layer, 22, 12, 32),
LRNLayerTest.Forward(layer, 32, 22, 42),
LRNLayerTest.Forward(layer, 42, 32),
});
Tensor dyTemp = new Tensor(null, expectedTemp.Shape);
dyTemp.Set(new float[]
{
1, 11, 21, 31, 41, 2, 12, 22, 32, 42
});
Tensor expectedDxTemp = new Tensor(null, xTemp.Shape);
expectedDxTemp.Set(new float[]
{
0.591914058f, 6.406341f, 11.8103943f, 16.4343262f, 22.1168289f, 1.18269f, 6.97113f, 12.3151608f, 16.8460541f, 22.5372047f
});
for (int i = 1; i <= 3; i++)
{
Session session = new Session();
Tensor x = session.Tile(xTemp, (int)Axis.B, i);
Tensor y = layer.Forward(session, new[] { x })[0];
Tensor expected = session.Tile(expectedTemp, (int)Axis.B, i);
Helpers.AreTensorsEqual(expected, y);
// unroll the graph
y.SetGradient(session.Tile(dyTemp, (int)Axis.B, i).Weights);
session.Unroll();
Tensor expectedDx = session.Tile(expectedDxTemp, (int)Axis.B, i);
Helpers.AreArraysEqual(expectedDx.Length, expectedDx.Weights, x.Gradient);
}
}
