public void ComplexLearnTest()
{
var network = new Network();
var inLayer = network.NewLayer("in")
.Add(3, typeof(AnyNeuron));
var midLayer1 = network.NewLayer("mid")
.Add(3, typeof(AllWithBlockingNeuron))
.Add(3, typeof(AnyWithBlockingNeuron));
var outLayer = network.NewLayer("out")
.Add(3, typeof(AnyWithBlockingNeuron))
;
var testcase = new ComplexLearnCase();
inLayer.ConnetToLayer(midLayer1, network, null, false);
inLayer.ConnetToLayer(midLayer1, network, null, true);
midLayer1.ConnetToLayer(outLayer, network, null, false);
midLayer1.ConnetToLayer(outLayer, network, null, true);
int i = 0;
while (network.SumOfEasinessToLearn() > 0.00001)
{
i++;
if (i > 100000)
{
break;
}
network.StudyIteration(testcase, 2, 1);
}
for (int ii = 0; ii < 100; ii++)
{
network.ClearStates();
testcase.setRandomInitState(network);
network.Tick(0.1);
network.Tick(0.1);
var result = testcase.Test(network);
Assert.IsTrue(result == TestCase.ResultType.OkNegative || result == TestCase.ResultType.OkPositive, "Test " + ii);
}
}
public void ComplexLearnWithPredefinedNetworkTest()
{
var testcase = new ComplexLearnCase();
var network = new Network();
var inLayer = network.NewLayer("in")
.Add(5, typeof(AnyNeuron));
var midLayer1 = network.NewLayer("mid")
.Add(3, typeof(AllWithBlockingNeuron));
var outLayer = network.NewLayer("out")
.Add(4, typeof(AnyWithBlockingNeuron))
.Add(1, typeof(AllWithBlockingNeuron));
var random = new Random();
inLayer.Neurons[0].ConnectTo(midLayer1.Neurons[0], random.NextDouble(), false);
inLayer.Neurons[1].ConnectTo(midLayer1.Neurons[0], random.NextDouble(), false);
inLayer.Neurons[2].ConnectTo(midLayer1.Neurons[0], random.NextDouble(), false);
inLayer.Neurons[3].ConnectTo(midLayer1.Neurons[0], random.NextDouble(), false);
inLayer.Neurons[0].ConnectTo(midLayer1.Neurons[1], random.NextDouble(), false);
inLayer.Neurons[1].ConnectTo(midLayer1.Neurons[1], random.NextDouble(), false);
inLayer.Neurons[2].ConnectTo(midLayer1.Neurons[1], random.NextDouble(), false);
inLayer.Neurons[0].ConnectTo(midLayer1.Neurons[2], random.NextDouble(), false);
inLayer.Neurons[1].ConnectTo(midLayer1.Neurons[2], random.NextDouble(), false);
inLayer.Neurons[0].ConnectTo(outLayer.Neurons[0], random.NextDouble(), false);
inLayer.Neurons[1].ConnectTo(outLayer.Neurons[0], random.NextDouble(), true);
inLayer.Neurons[2].ConnectTo(outLayer.Neurons[0], random.NextDouble(), true);
inLayer.Neurons[3].ConnectTo(outLayer.Neurons[0], random.NextDouble(), true);
inLayer.Neurons[4].ConnectTo(outLayer.Neurons[0], random.NextDouble(), true);
midLayer1.Neurons[2].ConnectTo(outLayer.Neurons[1], random.NextDouble(), false);
inLayer.Neurons[2].ConnectTo(outLayer.Neurons[1], random.NextDouble(), true);
inLayer.Neurons[3].ConnectTo(outLayer.Neurons[1], random.NextDouble(), true);
inLayer.Neurons[4].ConnectTo(outLayer.Neurons[1], random.NextDouble(), true);
midLayer1.Neurons[1].ConnectTo(outLayer.Neurons[2], random.NextDouble(), false);
inLayer.Neurons[3].ConnectTo(outLayer.Neurons[2], random.NextDouble(), true);
inLayer.Neurons[4].ConnectTo(outLayer.Neurons[2], random.NextDouble(), true);
midLayer1.Neurons[0].ConnectTo(outLayer.Neurons[3], random.NextDouble(), false);
inLayer.Neurons[4].ConnectTo(outLayer.Neurons[3], random.NextDouble(), true);
inLayer.Neurons[0].ConnectTo(outLayer.Neurons[4], random.NextDouble(), false);
inLayer.Neurons[1].ConnectTo(outLayer.Neurons[4], random.NextDouble(), false);
inLayer.Neurons[2].ConnectTo(outLayer.Neurons[4], random.NextDouble(), false);
inLayer.Neurons[3].ConnectTo(outLayer.Neurons[4], random.NextDouble(), false);
inLayer.Neurons[4].ConnectTo(outLayer.Neurons[4], random.NextDouble(), false);
int i = 0;
while (network.SumOfEasinessToLearn() > 0.00001)
{
i++;
if (i > 100000)
{
break;
}
network.StudyIteration(testcase, 2, 1);
}
for (int ii = 0; ii < 100; ii++)
{
network.ClearStates();
testcase.setRandomInitState(network);
network.Tick(0.1);
network.Tick(0.1);
var result = testcase.Test(network);
Assert.IsTrue(result == TestCase.ResultType.OkNegative || result == TestCase.ResultType.OkPositive, "Test " + ii);
}
}