public Network Train(Network net)
{
Console.WriteLine("=======Training Started========");
var iterationError = new double[net.InputLayer.Trainset.Length];
double errorRate;
do
{
for (var i = 0; i < net.InputLayer.Trainset.Length; ++i)
{
net.HiddenLayer.SetData(net.InputLayer.Trainset[i].Item1.ToDoubles());
net.HiddenLayer.Recognize(null, net.OutputLayer);
net.OutputLayer.Recognize(net, null);
var errors = new double[net.InputLayer.Trainset[i].Item2.OpCount];
for (var x = 0; x < errors.Length; ++x)
{
errors[x] = net.InputLayer.Trainset[i].Item2[x] - net.FactResult[x];
}
iterationError[i] = ErrorCalculator.CalcIterationError(errors);
var tempGsums = net.OutputLayer.BackwardPass(errors);
net.HiddenLayer.BackwardPass(tempGsums);
}
errorRate = ErrorCalculator.CalcRoundError(iterationError);
Console.WriteLine($"Round error: {errorRate}");
} while (errorRate > _allowedErrorRate);
net.HiddenLayer.InitWeights(MemoryModes.Set);
net.OutputLayer.InitWeights(MemoryModes.Set);
Console.WriteLine("========Training Ended========");
return(net);
}
