public static void GetAllWeidtg()
{
var tmp = new np4load(@"..\..\..\NeuralNetworks\Cardiology\инфаркт_миокарда.np4");
var neuralNet = tmp.GetNeuralNetwork();
var w = tmp.GetWeigth();
neuralNet.DecodeFromArray(w);
}
private void cbTrainsLog_SelectionChanged(object sender, SelectionChangedEventArgs e)
{
if (cbTrainsLog.SelectedItem != null)
{
InitEvent ie = cbTrainsLog.SelectedItem as InitEvent;
//ist<int> ageNum = new List<int>(ie.IterationCount);
//int[] ageNum2 = new int[ie.IterationCount];
//Parallel.For(1, ie.IterationCount+1, i => ageNum2[i-1] = i);
//for (int i = 1; i <= ie.IterationCount; i++)
// ageNum.Add(i);
iterations = rep.GetIteration(ie.Sid).OrderBy(x => x.AgeNumber);
cbAgeNumber.ItemsSource = iterations;
currentXmlNet = new np4load(ie.Path);
neuralNetIter = currentXmlNet.GetNeuralNetwork();
trainingData = currentXmlNet.GetTrainingData();
graphIterations = null;
testD.ClearAll();
pairDiagram.ClearAll();
}
}
public static void TrainNet()
{
var tmp = new np4load(@"..\..\..\NeuralNetworks\Cardiology\инфаркт_миокарда.np4");
TrainLogger tLog = TrainLogger.GetTrainLogger();
Dictionary<string, object> dLog = new Dictionary<string, object>();
var stopParam = tmp.GetTrainingStopParams();
int iterCount = 2500;//13943;
dLog.Add("event", "init");
dLog.Add("NeuralNetName", "инфаркт_миокарда.np4");
dLog.Add("Path", @"..\..\..\NeuralNetworks\Cardiology\инфаркт_миокарда.np4");
dLog.Add("Анкета", "Кардиология");
dLog.Add("IterationCount", iterCount);
dLog.Add("Time", DateTime.Now);
BasicMLDataSet tData = (BasicMLDataSet) tmp.GetTrainingData();
dLog.Add("TrainDataSize", tData.Count);
//записать входные данные и проскалированные
tLog.WriteEvent(dLog);
BasicNetwork neuralNet = tmp.GetNeuralNetwork();
MediatorNsimLinearScale scale = new MediatorNsimLinearScale();
DataProcessorConf dp = tmp.GetDataProcessor();
scale.DataProcessorM(dp);
//SearchHashForm search = new SearchHashForm();
//Guid[] gidForm = search.GetGIDForm(tData);
BasicMLDataSet tDataScale = scale.ProcessDataSet(tData);
IMLTrain trainMetod;
switch (tmp.GetNameTrainMetod())
{
case "ResilientPropagation":
{
trainMetod = new ResilientPropagation(neuralNet, tDataScale);
break;
}
default:
{
trainMetod = new ResilientPropagation(neuralNet, tDataScale);
break;
}
}
//trainMetod.Iteration(stopParam.Iterations);
trainMetod.Iteration(iterCount);
//var arrWeight = neuralNet.DumpWeights().Split(',');
//using (FileLogger fl = FileLogger.GetLogger())
//{
// fl.WriteString(neuralNet.DumpWeights());
// for (int i = 0; i < tDataScale.Data.Count; i++)//var item in tDataScale.Data)
// {
// fl.WriteString("Входной вектор: " + GetString(tData.Data[i].InputArray));
// var res = neuralNet.Compute(tDataScale.Data[i].Input);
// fl.WriteString("Результат: " + GetString(res.Data));
// var sRes = scale.RestoreIdealVector(res);
// fl.WriteString("Результат востановленный: " + GetString(sRes.Data));
// fl.WriteString("\n");
// }
//}
trainMetod.FinishTraining();
//int countW = neuralNet.EncodedArrayLength();
//double[] w = new double[countW];
//neuralNet.EncodeToArray(w);
Console.WriteLine("!!!");
//for (int i = 0; i < stopParam.Iterations; i++)
//{
// trainMetod.Iteration(
//}
}