public void TestDynamicXOR()
{
Func <int, int, double> inputFunc = OnInputFunc;
Func <int, int, double> idealFunc = delegate(int chunk, int index) { return(XOR.XORIdeal[chunk][index]); };
var input = new FuncMLDataProvider(inputFunc, XOR.XORInput.Length, XOR.XORInput[0].Length);
var ideal = new FuncMLDataProvider(idealFunc, XOR.XORIdeal.Length, XOR.XORIdeal[0].Length);
var ds = new DynamicMLDataSet(input, ideal);
var network = new BasicNetwork();
network.AddLayer(new BasicLayer(new ActivationSigmoid(), false, ds.InputSize));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), false, ds.InputSize + 1));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), false, ds.IdealSize));
network.Structure.FinalizeStructure();
network.Reset(42);
var trainer = new Encog.Neural.Networks.Training.Propagation.Resilient.ResilientPropagation(network, ds);
int maxIteration = 300;
int iteration = 0;
do
{
trainer.Iteration();
Debug.WriteLine(++iteration + ": Error = " + trainer.Error);
} while(trainer.Error > 0.0001 && maxIteration > iteration);
Assert.IsTrue(iteration < maxIteration);
}
public void BasicSlidingSineSignal()
{
var listSize = 30 * 200;
var inputList = new List <double>(listSize);
var idealList = new List <double>(listSize);
var rand = new Random(23);
for (int i = 0; i < listSize; i++)
{
idealList.Add(Math.Sin(Math.PI * 2.0 * i / 30));
inputList.Add(idealList[idealList.Count - 1] + (rand.NextDouble() - 0.5) * 0.1);
}
var input = new SlidingWindowMLDataProvider(inputList, 10, 0, 1);
var ideal = new SlidingWindowMLDataProvider(idealList, 2, 11, 1); // predecit the eleventh, twelth item from the ten previous to it
var ds = new DynamicMLDataSet(input, ideal);
Assert.AreEqual(10, input.WindowSize);
Assert.AreEqual(10, ds.InputSize);
Assert.AreEqual(2, ds.IdealSize);
Assert.AreEqual(listSize, ds.Count);
var network = new BasicNetwork();
network.AddLayer(new BasicLayer(ds.InputSize));
network.AddLayer(new BasicLayer(ds.InputSize + 3));
network.AddLayer(new BasicLayer(ds.IdealSize));
network.Structure.FinalizeStructure();
network.Reset(42);
var trainer = new Encog.Neural.Networks.Training.Propagation.Resilient.ResilientPropagation(network, ds);
int maxIteration = 300;
int iteration = 0;
do
{
trainer.Iteration();
Debug.WriteLine(++iteration + ": Error = " + trainer.Error);
} while(trainer.Error > 0.001 && maxIteration > iteration);
Assert.IsTrue(iteration < maxIteration);
}
public void BasicSlidingSineSignal()
{
var listSize = 30 * 200;
var inputList = new List<double>(listSize);
var idealList = new List<double>(listSize);
var rand = new Random(23);
for(int i = 0; i < listSize; i++)
{
idealList.Add(Math.Sin(Math.PI * 2.0 * i / 30));
inputList.Add(idealList[idealList.Count - 1] + (rand.NextDouble() - 0.5) * 0.1);
}
var input = new SlidingWindowMLDataProvider(inputList, 10, 0, 1);
var ideal = new SlidingWindowMLDataProvider(idealList, 2, 11, 1); // predecit the eleventh, twelth item from the ten previous to it
var ds = new DynamicMLDataSet(input, ideal);
Assert.AreEqual(10, input.WindowSize);
Assert.AreEqual(10, ds.InputSize);
Assert.AreEqual(2, ds.IdealSize);
Assert.AreEqual(listSize, ds.Count);
var network = new BasicNetwork();
network.AddLayer(new BasicLayer(ds.InputSize));
network.AddLayer(new BasicLayer(ds.InputSize + 3));
network.AddLayer(new BasicLayer(ds.IdealSize));
network.Structure.FinalizeStructure();
network.Reset(42);
var trainer = new Encog.Neural.Networks.Training.Propagation.Resilient.ResilientPropagation(network, ds);
int maxIteration = 300;
int iteration = 0;
do
{
trainer.Iteration();
Debug.WriteLine(++iteration + ": Error = " + trainer.Error);
} while(trainer.Error > 0.001 && maxIteration > iteration);
Assert.IsTrue(iteration < maxIteration);
}
public void TestDynamicXOR()
{
Func<int, int, double> inputFunc = OnInputFunc;
Func<int, int, double> idealFunc = delegate(int chunk, int index) { return XOR.XORIdeal[chunk][index]; };
var input = new FuncMLDataProvider(inputFunc, XOR.XORInput.Length, XOR.XORInput[0].Length);
var ideal = new FuncMLDataProvider(idealFunc, XOR.XORIdeal.Length, XOR.XORIdeal[0].Length);
var ds = new DynamicMLDataSet(input, ideal);
var network = new BasicNetwork();
network.AddLayer(new BasicLayer(new ActivationSigmoid(), false, ds.InputSize));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), false, ds.InputSize + 5));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), false, ds.IdealSize));
network.Structure.FinalizeStructure();
network.Reset(42);
var trainer = new Encog.Neural.Networks.Training.Propagation.Resilient.ResilientPropagation(network, ds);
int maxIteration = 300;
int iteration = 0;
do
{
trainer.Iteration();
Debug.WriteLine(++iteration + ": Error = " + trainer.Error);
} while(trainer.Error > 0.0001 && maxIteration > iteration);
Assert.IsTrue(iteration < maxIteration);
}
