public NodeFactory(NodeFuncGenerator funcGenerator, IDictionary <NodeType, double> probabilityMap)
{
if (probabilityMap.Count == 0)
{
throw new ArgumentException();
}
foreach (var item in probabilityMap)
{
if (item.Value < 0)
{
throw new ArgumentException();
}
}
funcGenerator_ = funcGenerator;
typeDistributionMap_ = new Dictionary <NodeType, double>();
double sum = 0.0d;
for (int i = 0; i < probabilityMap.Count; i++)
{
sum += probabilityMap.ElementAt(i).Value;
typeDistributionMap_.Add(probabilityMap.ElementAt(i).Key, sum);
}
typeDistributionMapMax_ = sum;
}
public TeacherTester()
{
Logger.Log(this, "App start");
NodeFuncGenerator funcGenerator = new NodeFuncGenerator();
Dictionary <NodeType, double> nodeProbabilityMap = new Dictionary <NodeType, double>()
{
{ NodeType.add, 1.0d },
{ NodeType.compare, 1.0d },
{ NodeType.copy, 0.0d },
{ NodeType.inverse, 1.0d },
{ NodeType.log, 1.0d },
{ NodeType.multiply, 1.0d },
{ NodeType.reverse, 1.0d }
};
NodeFactory nodeFactory = new NodeFactory(funcGenerator, nodeProbabilityMap);
List <DataPoint> dataPoints = GenerateSinWaveDataPoints(100, 0.19d);
Network network = new Network(nodeFactory, dataPoints, 1000);
double addRemoveChance = 0.5;
PatchFactory patchFactory = new PatchFactory(network, nodeFactory, addRemoveChance);
NetworkTeacher networkTeacher = new NetworkTeacher(network, patchFactory);
Logger.Log(this, String.Format("initial error: {0}", network.CalculateTrainingSqrError()));
networkTeacher.Teach(
teachingTime: TimeSpan.FromSeconds(10000000.0d),
wantedNoOfNodes: 300,
changesPerPatch: 8,
patchesPerTeachingSerie: 1
);
Logger.Log(this, String.Format("error after learning: {0}", network.CalculateTrainingSqrError()));
Logger.Log(this, "App finish");
Console.ReadKey();
}
public TeacherTester()
{
Logger.Log(this, "App start");
NodeFuncGenerator funcGenerator = new NodeFuncGenerator();
Dictionary<NodeType, double> nodeProbabilityMap = new Dictionary<NodeType, double>()
{
{NodeType.add, 1.0d},
{NodeType.compare, 1.0d},
{NodeType.copy, 0.0d},
{NodeType.inverse, 1.0d},
{NodeType.log, 1.0d},
{NodeType.multiply, 1.0d},
{NodeType.reverse, 1.0d}
};
NodeFactory nodeFactory = new NodeFactory(funcGenerator, nodeProbabilityMap);
List<DataPoint> dataPoints = GenerateSinWaveDataPoints(100, 0.19d);
Network network = new Network(nodeFactory, dataPoints, 1000);
double addRemoveChance = 0.5;
PatchFactory patchFactory = new PatchFactory(network, nodeFactory, addRemoveChance);
NetworkTeacher networkTeacher = new NetworkTeacher(network, patchFactory);
Logger.Log(this, String.Format("initial error: {0}", network.CalculateTrainingSqrError()));
networkTeacher.Teach(
teachingTime: TimeSpan.FromSeconds(10000000.0d),
wantedNoOfNodes: 300,
changesPerPatch: 8,
patchesPerTeachingSerie: 1
);
Logger.Log(this, String.Format("error after learning: {0}", network.CalculateTrainingSqrError()));
Logger.Log(this, "App finish");
Console.ReadKey();
}
public void Transform(NodeType nodeType)
{
nodeType_ = nodeType;
ParentsNo = 2;
if (nodeType == TrendPredictorLib.NodeType.copy) //IMPORTANT: TODO: This is heavy hack - inputs was not working - fix it by changing this property someway for inputs
{
ParentsNo = 1;
}
calculateFoo_ = NodeFuncGenerator.GenerateFunc(nodeType);
}
public void SetUp()
{
NodeFuncGenerator nodeFuncGenerator = new NodeFuncGenerator();
NodeFactory nodeFactory = new NodeFactory(
nodeFuncGenerator,
new Dictionary<NodeType, double>() { { NodeType.add, 1.0d } });
network = new Network(
nodeFactory: nodeFactory,
trainingData: new List<DataPoint>()
{
new DataPoint(new List<double>(){2.0d, 2.0d, 2.0d}, 5.0d) //IMPORTANT: TODO: was modified from "new DataPoint(new List<double>(){2.0d, 2.0d}, 5.0d)" - now test can work incorrectly
},
maxHierarchy: 1000
);
patchFactory = new PatchFactory(network, nodeFactory, 1.0d);
networkTeacher = new NetworkTeacher(network, patchFactory);
}
public NodeFactory(NodeFuncGenerator funcGenerator, IDictionary<NodeType, double> probabilityMap)
{
if (probabilityMap.Count == 0)
throw new ArgumentException();
foreach (var item in probabilityMap)
{
if (item.Value < 0)
throw new ArgumentException();
}
funcGenerator_ = funcGenerator;
typeDistributionMap_ = new Dictionary<NodeType, double>();
double sum = 0.0d;
for (int i = 0; i < probabilityMap.Count; i++)
{
sum += probabilityMap.ElementAt(i).Value;
typeDistributionMap_.Add(probabilityMap.ElementAt(i).Key, sum);
}
typeDistributionMapMax_ = sum;
}
public void SinWaveTrainingTest()
{
NodeFuncGenerator funcGenerator = new NodeFuncGenerator();
Dictionary<NodeType, double> nodeProbabilityMap = new Dictionary<NodeType,double>()
{
{NodeType.add, 1.0d},
{NodeType.compare, 1.0d},
{NodeType.copy, 0.0d},
{NodeType.inverse, 1.0d},
{NodeType.log, 1.0d},
{NodeType.multiply, 1.0d},
{NodeType.reverse, 1.0d}
};
NodeFactory nodeFactory = new NodeFactory(funcGenerator, nodeProbabilityMap);
List<DataPoint> dataPoints = GenerateSinWaveDataPoints(100, 0.35d);
Network network = new Network(nodeFactory, dataPoints, 1000);
double addRemoveChance = 0.5;
PatchFactory patchFactory = new PatchFactory(network, nodeFactory, addRemoveChance);
NetworkTeacher networkTeacher = new NetworkTeacher(network, patchFactory);
networkTeacher.Teach(TimeSpan.FromSeconds(0.1d), 2000, 3, 30);
}
