private static void setIndexesToSheet(Subnet subnet)
{
int count = ANNUtils.getCountOfWeights(subnet.Network);
for (int i = 1; i <= count; i++)
{
mediumsSheet.Cells[ i + 1, 1].Value = "Вес " + (i-1).ToString();
coefvariationsSheet.Cells[i + 1, 1].Value = "Вес " + (i - 1).ToString();
deviationSheet.Cells[i + 1, 1].Value = "Вес " + (i - 1).ToString();
}
}
public virtual double[] Compute(double[] input, Subnet subnet, int numberLayer, List<Record> MinMaxValues)
{
// local variable to avoid mutlithread conflicts
double[] output = new double[neuronsCount];
// compute each neuron
for (int i = 0; i < neurons.Length; i++)
{
output[i] = neurons[i].Compute(input, subnet, numberLayer, i, MinMaxValues);
}
// assign output property as well (works correctly for single threaded usage)
this.output = output;
return output;
}
//запись весов отобранных или скрещенных сетей
public static void SubnetToFile(Subnet net, int reconnectingCount, Boolean cross, string parents = "", int population = 0)
{
try
{
string folder = "";
if (cross)
folder = currentFolder + "\\" + CROSS_DIR + reconnectingCount.ToString() + "\\" + population + "\\";
else
folder = currentFolder + "\\" + SELECTION_DIR + reconnectingCount.ToString() + "\\";
LogHelper.checkDir(folder);
string filename = Path.Combine(folder, net.ID + "-" + parents + "-" + net.quality + ".csv");
System.IO.StreamWriter file = new System.IO.StreamWriter(filename);
file.WriteLine("Layer;Neuron;Weight;Value;");
LogHelper.allWeightsSheet.Cells[ 1, netsCounter + 1].Value = net.ID + "-" + parents + "-" + net.quality;
int row = 0;
for (int i = 0; i < net.Network.Layers.Length; i++)
{
for (int j = 0; j < net.Network.Layers[i].Neurons.Length; j++)
{
for (int k = 0; k < net.Network.Layers[i].Neurons[j].Weights.Length; k++)
{
file.WriteLine(i.ToString() + ";" + j.ToString() + ";" + k.ToString() + ";" + net.Network.Layers[i].Neurons[j].Weights[k] + ";");
LogHelper.allWeightsSheet.Cells[row + 2, netsCounter + 1].Value = net.Network.Layers[i].Neurons[j].Weights[k];
row++;
}
}
}
file.Flush();
file.Close();
netsCounter++;
}
catch
{
// Перехватываем все и ничего не делаем
}
}
//Инициализирует папки, файлы и spreadsheets
public static void InitReporting(Subnet subnet = null, int reconnectingCount = 0)
{
LogHelper.checkDir(LogHelper.currentFolder);
LogHelper.netsCounter = 0;
//создание файла с результатами корректировки
LogHelper.resultsFile = new FileInfo(LogHelper.currentFolder + "\\" + RESULT_CORRECT_FILE + reconnectingCount.ToString() + ".xlsx");
LogHelper.results = new ExcelPackage(LogHelper.resultsFile);
LogHelper.resultsSheet = LogHelper.results.Workbook.Worksheets.Add("Результаты");
//создание файла с коэф.вариации
LogHelper.covarFile = new FileInfo(LogHelper.currentFolder + "\\" + VARIATIONS_FILE + reconnectingCount.ToString() + ".xlsx");
LogHelper.coefvariations = new ExcelPackage(LogHelper.covarFile);
LogHelper.coefvariationsSheet = LogHelper.coefvariations.Workbook.Worksheets.Add("Коэф.вариации");
//стандартное отклонение
LogHelper.deviationFile = new FileInfo(LogHelper.currentFolder + "\\" + DEVIATION_FILE + reconnectingCount.ToString() + ".xlsx");
LogHelper.deviationBook = new ExcelPackage(LogHelper.deviationFile);
LogHelper.deviationSheet = LogHelper.deviationBook.Workbook.Worksheets.Add("Стандартные отклонения весов");
//средние значения
LogHelper.mediumsFile = new FileInfo(LogHelper.currentFolder + "\\" + MEDIUMS_FILE + reconnectingCount.ToString() + ".xlsx");
LogHelper.mediumsBook = new ExcelPackage(LogHelper.mediumsFile);
LogHelper.mediumsSheet = LogHelper.mediumsBook.Workbook.Worksheets.Add("Средние значения");
//веса всех сетей
LogHelper.allWeightsFile = new FileInfo(LogHelper.currentFolder + "\\" + ALLWEIGHTS_FILE + reconnectingCount.ToString() + ".xlsx");
LogHelper.allWeightsBook = new ExcelPackage(LogHelper.allWeightsFile);
LogHelper.allWeightsSheet = LogHelper.allWeightsBook.Workbook.Worksheets.Add("Веса всех сетей сессии");
//параметры коррекции по поколениям
LogHelper.correctParamsFile = new FileInfo(LogHelper.currentFolder + "\\" + PARAMS_CORRECT_FILE + reconnectingCount.ToString() + ".xlsx");
LogHelper.correctParamsBook = new ExcelPackage(LogHelper.correctParamsFile);
LogHelper.correctParamsSheet = LogHelper.correctParamsBook.Workbook.Worksheets.Add("Показатели коррекции по поколениям");
LogHelper.correctParamsSheet = LogHelper.setCommonCorrectParamsSheetHeaders(correctParamsSheet);
if (subnet != null)
LogHelper.setIndexesToSheet(subnet);
return;
}
public abstract double Compute(double[] input, Subnet subnet, int numberLayer, int numberNeuron, List<Record> MinMaxValues);
public override double Compute(double[] input, Subnet subnet, int numberLayer, int numberNeuron, List<Record> MinMaxValues)
{
Random rnd = new Random();
// check for corrent input vector
if (input.Length != inputsCount)
throw new ArgumentException("Wrong length of the input vector.");
// initial sum value
double sum = 0.0;
// compute weighted sum of inputs
for (int i = 0; i < weights.Length; i++)
{
if (subnet.outputAssosiated.Contains(numberLayer.ToString() + ":" + numberNeuron.ToString() + ":" + i.ToString()))
{
double min = 0.0;
double max = 0.0;
foreach (Record elem in MinMaxValues)
{
if ((elem.numberLayer == numberLayer) && (elem.numberNeuron == numberNeuron) && (elem.numberWeight == i))
{
min = elem.min;
max = elem.max;
}
}
double normValue = (((subnet.res[0] - (-1)) * (max - min)) / (1 - (-1))) + min;
sum += (weights[i] * normValue ) * input[i];
// weights[i] = weights[i] * subnet.res[0];
subnet.res.RemoveAt(0);
}
else sum += weights[i] * input[i];
}
sum += threshold;
// local variable to avoid mutlithreaded conflicts
double output = function.Function(sum);
// assign output property as well (works correctly for single threaded usage)
this.output = output;
return output;
}
//compute net with subnets
public virtual double[] Compute(double[] input, Subnet subnet, List<Record> MinMaxValues)
{
List<String> inputs = subnet.inputAssosiated;
double[] weights = this.getWeights(inputs, input);
subnet.Compute(weights);
// local variable to avoid mutlithread conflicts
double[] output = input;
// compute each layer
for (int i = 0; i < layers.Length; i++)
{
output = layers[i].Compute(output, subnet, i, MinMaxValues);
}
// assign output property as well (works correctly for single threaded usage)
this.output = output;
return output;
}