public SigmoidNeuron(string NeuronData, NeuronLayer PreviousLayer)
{
if (PreviousLayer != null)
{
string[] DataSplit = NeuronData.Split('|');
string[] WeightsData = DataSplit[0].Split(',');
Bias = float.Parse(DataSplit[1]);
Weights = new float[WeightsData.Length];
Inputs = new SigmoidNeuron[WeightsData.Length];
NablaWeights = new float[PreviousLayer.Nodes.Length];
for (int i = 0; i < WeightsData.Length; ++i)
{
Weights[i] = float.Parse(WeightsData[i]);
Inputs[i] = PreviousLayer.Nodes[i];
NablaWeights[i] = 0;
}
}
else
{
Bias = 0;
}
}
public NeuronLayer(int NumNodes, NeuronLayer PreviousLayer)
{
Nodes = new SigmoidNeuron[NumNodes];
for (int i = 0; i < NumNodes; ++i)
{
Nodes[i] = new SigmoidNeuron(PreviousLayer);
}
}
public NeuronLayer(int NumNodes, string[] NeuronData, NeuronLayer PreviousLayer)
{
Nodes = new SigmoidNeuron[NumNodes];
for (int i = 0; i < NumNodes; ++i)
{
Nodes[i] = new SigmoidNeuron((NeuronData.Length > 0) ? NeuronData[i] : null, PreviousLayer);
}
}
public void SetupNet()
{
// if there is a pre-exsiting net layout use that. otherwise build a new net.
if (NetLayout == null)
{
NetLayers = new NeuronLayer[NumLayers];
for (int i = 0; i < NumLayers; ++i)
{
NetLayers[i] = new NeuronLayer(LayerNeuronCounts[i], (i > 0) ? NetLayers[i - 1] : null);
}
Outputs = new float[LayerNeuronCounts[NumLayers - 1]];
ExpectedOutputs = new float[LayerNeuronCounts[NumLayers - 1]];
}
else
{
LoadNetFromFile();
}
}
void LoadNetFromFile()
{
string[] Neurons = NetLayout.text.Split('\n');
NetLayers = new NeuronLayer[NumLayers];
int StartCount = 0;
for (int i = 0; i < NumLayers; ++i)
{
string[] LayerNeuronData = GetNeuronRange(StartCount, (i > 0) ? LayerNeuronCounts[i] + StartCount : 0, Neurons);
NetLayers[i] = new NeuronLayer(LayerNeuronCounts[i], LayerNeuronData, (i > 0) ? NetLayers[i - 1] : null);
StartCount += (i > 0) ? LayerNeuronCounts[i] : 0;
}
Outputs = new float[LayerNeuronCounts[NumLayers - 1]];
ExpectedOutputs = new float[LayerNeuronCounts[NumLayers - 1]];
}
public SigmoidNeuron(NeuronLayer PreviousLayer)
{
if (PreviousLayer != null)
{
Weights = new float[PreviousLayer.Nodes.Length];
Inputs = new SigmoidNeuron[PreviousLayer.Nodes.Length];
NablaWeights = new float[PreviousLayer.Nodes.Length];
for (int i = 0; i < PreviousLayer.Nodes.Length; ++i)
{
Inputs[i] = PreviousLayer.Nodes[i];
Weights[i] = Random.Range(-1f, 1f);
NablaWeights[i] = 0;
}
Bias = Random.Range(-1f, 1f);
}
else
{
Bias = 0;
}
}
public SigmoidNeuron(NeuronLayer PreviousLayer)
{
if (PreviousLayer != null)
{
System.Random myRand = new System.Random();
Weights = new float[PreviousLayer.Nodes.Length];
Inputs = new SigmoidNeuron[PreviousLayer.Nodes.Length];
NablaWeights = new float[PreviousLayer.Nodes.Length];
for (int i = 0; i < PreviousLayer.Nodes.Length; ++i)
{
Inputs[i] = PreviousLayer.Nodes[i];
Weights[i] = (float)((myRand.NextDouble() - 0.5) * 2);//Random.Range(-1f, 1f);
NablaWeights[i] = 0;
}
Bias = (float)((myRand.NextDouble() - 0.5) * 2);//Random.Range(-1f, 1f);
}
else
{
Bias = 0;
}
}
