/// <summary>
/// Load a vector from a file
/// </summary>
public static SimpleNeuralNetwork Load(BinaryReader r)
{
SimpleNeuralNetwork snn = new SimpleNeuralNetwork();
snn.inputsCount = r.ReadInt32();
snn.hiddenCount = r.ReadInt32();
snn.outputCount = r.ReadInt32();
snn.weights_IH = new double[snn.inputsCount, snn.hiddenCount];
snn.weights_HO = new double[snn.hiddenCount, snn.outputCount];
for (int i = 0; i < snn.inputsCount; i++)
{
for (int h = 0; h < snn.hiddenCount; h++)
{
snn.weights_IH[i, h] = r.ReadDouble();
}
}
for (int h = 0; h < snn.hiddenCount; h++)
{
for (int o = 0; o < snn.outputCount; o++)
{
snn.weights_HO[h, o] = r.ReadDouble();
}
}
snn.neurons_H = new double[r.ReadInt32()];
for (int h = 0; h < snn.hiddenCount; h++)
{
snn.neurons_H[h] = r.ReadDouble();
}
snn.neurons_O = DoubleVector.Load(r);
snn.error_H = new double[r.ReadInt32()];
for (int h = 0; h < snn.error_H.Length; h++)
{
snn.error_H[h] = r.ReadDouble();
}
snn.error_O = new double[r.ReadInt32()];
for (int h = 0; h < snn.error_O.Length; h++)
{
snn.error_O[h] = r.ReadDouble();
}
snn.learningRate = r.ReadDouble();
return(snn);
}
public SimpleNeuralNetSolver(string characterSet, int imageWidth, int imageHeight, int hiddenNeurons, int charactersPerImage, double learningRate)
{
// Split the set of characters into a list
charsSet = characterSet.ToCharStringList();
charsPerImage = charactersPerImage;
learnRate = learningRate;
ExpectedWidth = imageWidth;
ExpectedHeight = imageHeight;
sann = new SimpleNeuralNetwork(imageWidth * imageHeight, hiddenNeurons, characterSet.Length, learningRate);
sann.OnTrainingProgressChange += new SimpleNeuralNetwork.TrainingProgressHandler(sann_OnTrainingProgressChange);
trainerWorker = new BackgroundWorker();
trainerWorker.DoWork += new DoWorkEventHandler(worker_DoWork);
trainerWorker.RunWorkerCompleted += new RunWorkerCompletedEventHandler(worker_RunWorkerCompleted);
trainerWorker.ProgressChanged += new ProgressChangedEventHandler(worker_ProgressChanged);
trainerWorker.WorkerReportsProgress = true;
testerWorker = new BackgroundWorker();
testerWorker.DoWork += new DoWorkEventHandler(testerWorker_DoWork);
testerWorker.RunWorkerCompleted += new RunWorkerCompletedEventHandler(testerWorker_RunWorkerCompleted);
}
/// <summary>
/// Load a vector from a file
/// </summary>
public static SimpleNeuralNetwork Load(BinaryReader r)
{
SimpleNeuralNetwork snn = new SimpleNeuralNetwork();
snn.inputsCount = r.ReadInt32();
snn.hiddenCount = r.ReadInt32();
snn.outputCount = r.ReadInt32();
snn.weights_IH = new double[snn.inputsCount, snn.hiddenCount];
snn.weights_HO = new double[snn.hiddenCount, snn.outputCount];
for (int i = 0; i < snn.inputsCount; i++)
{
for (int h = 0; h < snn.hiddenCount; h++)
{
snn.weights_IH[i, h] = r.ReadDouble();
}
}
for (int h = 0; h < snn.hiddenCount; h++)
{
for (int o = 0; o < snn.outputCount; o++)
{
snn.weights_HO[h, o] = r.ReadDouble();
}
}
snn.neurons_H = new double[r.ReadInt32()];
for (int h = 0; h < snn.hiddenCount; h++)
{
snn.neurons_H[h] = r.ReadDouble();
}
snn.neurons_O = DoubleVector.Load(r);
snn.error_H = new double[r.ReadInt32()];
for (int h = 0; h < snn.error_H.Length; h++)
{
snn.error_H[h] = r.ReadDouble();
}
snn.error_O = new double[r.ReadInt32()];
for (int h = 0; h < snn.error_O.Length; h++)
{
snn.error_O[h] = r.ReadDouble();
}
snn.learningRate = r.ReadDouble();
return snn;
}
public override void Load(BinaryReader r)
{
base.Load(r);
charsPerImage = r.ReadInt32();
charsSet = r.ReadString().ToCharArray().ToList().Select(c => c.ToString()).ToList();
learnRate = r.ReadDouble();
sann = SimpleNeuralNetwork.Load(r);
sann.OnTrainingProgressChange += new SimpleNeuralNetwork.TrainingProgressHandler(sann_OnTrainingProgressChange);
}
