static void PrintWeights(FastCyclicNetwork network)
{
foreach (var conn in network.ConnectionArray)
{
Console.WriteLine("[{0}] -> [{1}]: {2:N4}", conn._srcNeuronIdx, conn._tgtNeuronIdx, conn._weight);
}
}
public static CFastCyclicNetwork Convert(FastCyclicNetwork fcn)
{
var cfcn = new CFastCyclicNetwork();
var numberOfNeurons = fcn._neuronActivationFnArray.Length;
// Copy connections
cfcn.Connections = new List <CConnection>(fcn._connectionArray.Length);
foreach (var c in fcn._connectionArray)
{
var cc = new CConnection
{
ToNeuronId = c._tgtNeuronIdx,
FromNeuronId = c._srcNeuronIdx,
Weight = c._weight
};
cfcn.Connections.Add(cc);
}
// Copy activation functions
cfcn.ActivationFunctions = new List <string>(numberOfNeurons);
foreach (var s in fcn._neuronActivationFnArray)
{
if (s == null)
{
cfcn.ActivationFunctions.Add("");
}
else
{
cfcn.ActivationFunctions.Add(s.FunctionId);
}
}
// Copy auxiliary arguments
cfcn.NeuronAuxArgs = new List <List <double> >(numberOfNeurons);
foreach (var aux in fcn._neuronAuxArgsArray)
{
if (aux == null)
{
cfcn.NeuronAuxArgs.Add(new List <double>());
}
else
{
cfcn.NeuronAuxArgs.Add(new List <double>(aux));
}
}
cfcn.NeuronCount = fcn._neuronCount;
cfcn.InputNeuronCount = fcn._inputNeuronCount;
cfcn.OutputNeuronCount = fcn._outputNeuronCount;
cfcn.TimestepsPerActivation = fcn._timestepsPerActivation;
return(cfcn);
}
private static void RunXor(FastCyclicNetwork network, double[][] inputs)
{
for (int i = 0; i < inputs.Length; i++)
{
network.ResetState();
// Convert the sensors into an input array for the network
for (int j = 0; j < inputs[i].Length; j++)
{
network.InputSignalArray[j] = inputs[i][j];
}
// Activate the network
network.Activate();
Console.WriteLine("[{0},{1}] -> {2:N4}", inputs[i][0], inputs[i][1], network.OutputSignalArray[0]);
}
}
public FitnessInfo Evaluate(FastCyclicNetwork phenome)
{
EvaluationCount++;
var fcn = FastCyclicNetworkAdapter.Convert(phenome);
ProtocolManager.Open();
var cFitness = ProtocolManager.Client.calculateXorPhenotypeFitness(fcn);
//ProtocolManager.Close(); adding this line implies a huge performance hit
var auxFitness = new AuxFitnessInfo[cFitness.AuxFitness.Count];
for (int i = 0; i < auxFitness.Length; i++)
{
auxFitness[i] = new AuxFitnessInfo(cFitness.AuxFitness[i].Name, cFitness.AuxFitness[i].Value);
}
var fitness = new FitnessInfo(cFitness.Fitness, auxFitness);
StopConditionSatisfied = cFitness.StopConditionSatisfied;
if (StopConditionSatisfied)
{
FitnessInfo fi = _xorBlackBoxEvaluator.Evaluate(phenome);
bool reallySatisfied = _xorBlackBoxEvaluator.StopConditionSatisfied;
Debug.Assert((fi._fitness >= 10) == _xorBlackBoxEvaluator.StopConditionSatisfied);
if (!reallySatisfied)
{
Console.Out.WriteLine("ERROR: really fitness is not calculated (" + fi._fitness + " versus " +
cFitness.Fitness + ")");
Console.Out.WriteLine(phenome.ToString());
//throw new Exception("Wheih :(");
}
else
{
Console.Out.WriteLine("Networks is really good");
}
}
// We do not use the NEAT genome decoder. The NEAT genome decoder would
// do the same as the following code, but provide us with an IBlackBox object.
// Because we pass on the object to our Java code, we need to be aware of its
// underlying structure. The additional layer of abstraction gets in the way.
return(fitness);
}
private static double[][] TrainXOR(int epochs, FastCyclicNetwork network)
{
double[][] inputs = new double[][] {
new double[] { 0, 0 },
new double[] { 1, 0 },
new double[] { 0, 1 },
new double[] { 1, 1 }
};
double[][] outputs = new double[][] {
new double[] { 0 },
new double[] { 1 },
new double[] { 1 },
new double[] { 0 }
};
for (int i = 0; i < epochs; i++)
{
network.Train(inputs[i % inputs.Length], outputs[i % outputs.Length]);
}
return(inputs);
}
public FitnessInfo Evaluate(IBlackBox box)
{
EvaluationCount++;
FastCyclicNetwork network = box as FastCyclicNetwork;
// Write them out to a file.
DeepBeliefNetworkBiaserIO.WriteOutputMatrix(network._connectionArray,
Constants.INPUT_SIZE + 1,
Constants.OUTPUT_SIZE);
// Run the python DBN script which will output
DeepBeliefNetworkBiaserIO.RunPythonScriptMultithreaded();
Tuple <double, double> fitnessPair = DeepBeliefNetworkBiaserIO.ReadFitness();
// Set stop flag when max fitness is attained.
if (!StopConditionSatisfied && fitnessPair.Item1 == MaxFitness)
{
StopConditionSatisfied = true;
}
return(new FitnessInfo(fitnessPair.Item1, fitnessPair.Item2));
}
/// <summary>
/// Evaluate the provided IBlackBox against the XOR problem domain and return its fitness score.
/// </summary>
public FitnessInfo Evaluate(IBlackBox box)
{
double fitness = 0;
double output;
double pass = 1.0;
ISignalArray inputArr = box.InputSignalArray;
ISignalArray outputArr = box.OutputSignalArray;
_evalCount++;
// Train the network
FastCyclicNetwork net = (FastCyclicNetwork)box;
//----- Test 0,0
box.ResetState();
// Set the input values
inputArr[0] = 0.0;
inputArr[1] = 0.0;
// Activate the black box.
box.Activate();
if (!box.IsStateValid)
{ // Any black box that gets itself into an invalid state is unlikely to be
// any good, so lets just bail out here.
return(FitnessInfo.Zero);
}
// Read output signal.
output = Math.Min(1, Math.Max(0, outputArr[0]));
var output0 = output;
Debug.Assert(output >= 0.0, "Unexpected negative output.");
// Calculate this test case's contribution to the overall fitness score.
//fitness += 1.0 - output; // Use this line to punish absolute error instead of squared error.
fitness += 1.0 - (output * output);
if (output > 0.5)
{
pass = 0.0;
}
//----- Test 1,1
// Reset any black box state from the previous test case.
box.ResetState();
// Set the input values
inputArr[0] = 1.0;
inputArr[1] = 1.0;
// Activate the black box.
box.Activate();
if (!box.IsStateValid)
{ // Any black box that gets itself into an invalid state is unlikely to be
// any good, so lets just bail out here.
return(FitnessInfo.Zero);
}
// Read output signal.
output = Math.Min(1, Math.Max(0, outputArr[0]));
var output1 = output;
Debug.Assert(output >= 0.0, "Unexpected negative output.");
// Calculate this test case's contribution to the overall fitness score.
//fitness += 1.0 - output; // Use this line to punish absolute error instead of squared error.
fitness += 1.0 - (output * output);
if (output > 0.5)
{
pass = 0.0;
}
//----- Test 0,1
// Reset any black box state from the previous test case.
box.ResetState();
// Set the input values
inputArr[0] = 0.0;
inputArr[1] = 1.0;
// Activate the black box.
box.Activate();
if (!box.IsStateValid)
{ // Any black box that gets itself into an invalid state is unlikely to be
// any good, so lets just bail out here.
return(FitnessInfo.Zero);
}
// Read output signal.
output = Math.Min(1, Math.Max(0, outputArr[0]));
var output2 = output;
Debug.Assert(output >= 0.0, "Unexpected negative output.");
// Calculate this test case's contribution to the overall fitness score.
// fitness += output; // Use this line to punish absolute error instead of squared error.
fitness += 1.0 - ((1.0 - output) * (1.0 - output));
if (output <= 0.5)
{
pass = 0.0;
}
//----- Test 1,0
// Reset any black box state from the previous test case.
box.ResetState();
// Set the input values
inputArr[0] = 1.0;
inputArr[1] = 0.0;
// Activate the black box.
box.Activate();
if (!box.IsStateValid)
{ // Any black box that gets itself into an invalid state is unlikely to be
// any good, so lets just bail out here.
return(FitnessInfo.Zero);
}
// Read output signal.
output = Math.Min(1, Math.Max(0, outputArr[0]));
var output3 = output;
Debug.Assert(output >= 0.0, "Unexpected negative output.");
// Calculate this test case's contribution to the overall fitness score.
// fitness += output; // Use this line to punish absolute error instead of squared error.
fitness += 1.0 - ((1.0 - output) * (1.0 - output));
if (output <= 0.5)
{
pass = 0.0;
}
// If all four outputs were correct, that is, all four were on the correct side of the
// threshold level - then we add 10 to the fitness.
fitness += pass * 10.0;
if (fitness > 3.6)
{
Console.WriteLine("Before Nodes={5} Connections={6} [0,0]={0:N3} [1,1]={1:N3} [0,1]={2:N3} [1,0]={3:N3} fitness={4:N3}", output0, output1, output2, output3, fitness, net.HiddenCount, net.ConnectionCount);
}
fitness = 0;
net.BackpropLearningRate = 0.1;
net.Momentum = 0.9;
double[][] exampleInputs = new double[][] {
new double[] { 0, 0 },
new double[] { 0, 1 },
new double[] { 1, 0 },
new double[] { 1, 1 }
};
double[][] exampleOutputs = new double[][] {
new double[] { 0 },
new double[] { 1 },
new double[] { 1 },
new double[] { 0 }
};
net.RandomizeWeights();
for (int i = 0; i < 4000; i++)
{
net.Train(exampleInputs[i % 4], exampleOutputs[i % 4]);
}
//----- Test 0,0
box.ResetState();
// Set the input values
inputArr[0] = 0.0;
inputArr[1] = 0.0;
// Activate the black box.
box.Activate();
if (!box.IsStateValid)
{ // Any black box that gets itself into an invalid state is unlikely to be
// any good, so lets just bail out here.
return(FitnessInfo.Zero);
}
// Read output signal.
output = Math.Min(1, Math.Max(0, outputArr[0]));
output0 = output;
Debug.Assert(output >= 0.0, "Unexpected negative output.");
// Calculate this test case's contribution to the overall fitness score.
//fitness += 1.0 - output; // Use this line to punish absolute error instead of squared error.
fitness += 1.0 - (output * output);
if (output > 0.5)
{
pass = 0.0;
}
//----- Test 1,1
// Reset any black box state from the previous test case.
box.ResetState();
// Set the input values
inputArr[0] = 1.0;
inputArr[1] = 1.0;
// Activate the black box.
box.Activate();
if (!box.IsStateValid)
{ // Any black box that gets itself into an invalid state is unlikely to be
// any good, so lets just bail out here.
return(FitnessInfo.Zero);
}
// Read output signal.
output = Math.Min(1, Math.Max(0, outputArr[0]));
output1 = output;
Debug.Assert(output >= 0.0, "Unexpected negative output.");
// Calculate this test case's contribution to the overall fitness score.
//fitness += 1.0 - output; // Use this line to punish absolute error instead of squared error.
fitness += 1.0 - (output * output);
if (output > 0.5)
{
pass = 0.0;
}
//----- Test 0,1
// Reset any black box state from the previous test case.
box.ResetState();
// Set the input values
inputArr[0] = 0.0;
inputArr[1] = 1.0;
// Activate the black box.
box.Activate();
if (!box.IsStateValid)
{ // Any black box that gets itself into an invalid state is unlikely to be
// any good, so lets just bail out here.
return(FitnessInfo.Zero);
}
// Read output signal.
output = Math.Min(1, Math.Max(0, outputArr[0]));
output2 = output;
Debug.Assert(output >= 0.0, "Unexpected negative output.");
// Calculate this test case's contribution to the overall fitness score.
// fitness += output; // Use this line to punish absolute error instead of squared error.
fitness += 1.0 - ((1.0 - output) * (1.0 - output));
if (output <= 0.5)
{
pass = 0.0;
}
//----- Test 1,0
// Reset any black box state from the previous test case.
box.ResetState();
// Set the input values
inputArr[0] = 1.0;
inputArr[1] = 0.0;
// Activate the black box.
box.Activate();
if (!box.IsStateValid)
{ // Any black box that gets itself into an invalid state is unlikely to be
// any good, so lets just bail out here.
return(FitnessInfo.Zero);
}
// Read output signal.
output = Math.Min(1, Math.Max(0, outputArr[0]));
output3 = output;
Debug.Assert(output >= 0.0, "Unexpected negative output.");
// Calculate this test case's contribution to the overall fitness score.
// fitness += output; // Use this line to punish absolute error instead of squared error.
fitness += 1.0 - ((1.0 - output) * (1.0 - output));
if (output <= 0.5)
{
pass = 0.0;
}
// If all four outputs were correct, that is, all four were on the correct side of the
// threshold level - then we add 10 to the fitness.
fitness += pass * 10.0;
if (fitness >= StopFitness)
{
_stopConditionSatisfied = true;
}
if (fitness > 3.6)
{
Console.WriteLine("After Nodes={5} Connections={6} [0,0]={0:N3} [1,1]={1:N3} [0,1]={2:N3} [1,0]={3:N3} fitness={4:N3}", output0, output1, output2, output3, fitness, net.HiddenCount, net.ConnectionCount);
}
return(new FitnessInfo(fitness, fitness));
}
