public void ProblemOne()
{
// Use randomly selected data point .txt files to set Train and Test Table.
int theNumberOfOutputs = 10;
theDtFileOps.TrainTable = theDtFileOps.ConvertTextToDataTable(786, "TrainTable.txt");
theDtFileOps.TestTable = theDtFileOps.ConvertTextToDataTable(786, "TestTable.txt");
// Indicate import status.
Console.WriteLine("Files imported successfully.");
// Initialize and train network.
Network cNetwork = theBackpropagation.InitializeNetwork(theDtFileOps.TrainTable.Columns.Count - 1, 100, theNumberOfOutputs);
theBackpropagation.InitializeMatrix();
theBackpropagation.TrainNetwork(cNetwork, theDtFileOps.TrainTable, 0.01, 200, theNumberOfOutputs);
Console.WriteLine();
theBackpropagation.DisplayMatrix(theBackpropagation.ActualExpectedTrainMatrix);
theBackpropagation.InitializeMatrix();
// Test predictions using trained network.
foreach (DataRow row in theDtFileOps.TestTable.Rows)
{
List <double> theRowAsList = theBackpropagation.ConvertDataRowToListDouble(row);
double aPrediction = theBackpropagation.Predict(cNetwork, theRowAsList);
Console.WriteLine("Actual={0}, Predicted={1}", theRowAsList[theRowAsList.Count - 1], aPrediction);
theBackpropagation.ActualExpectedTestMatrix[Convert.ToInt32(aPrediction), Convert.ToInt32(theRowAsList[theRowAsList.Count - 1])]++;
}
// Print Confusion matrix for Train and Test sets.
Console.WriteLine();
theBackpropagation.DisplayMatrix(theBackpropagation.ActualExpectedTestMatrix);
// Print EpochError values to file for plotting in other environment.
theDtFileOps.ConvertListToTextFile(theBackpropagation.EpochErrorList, "EpochError.txt");
// Create data table and provide designated number of columns.
DataTable hiddenNeurons = new DataTable();
for (int col = 0; col < 784; col++)
{
hiddenNeurons.Columns.Add(new DataColumn("Column" + (col + 1).ToString()));
}
// Add all weights for each neuron in the final network hidden layer to a data table.
foreach (var neuron in theBackpropagation.OtherFinalNetworkHiddenLayer.Neurons)
{
DataRow row = hiddenNeurons.NewRow();
row.ItemArray = neuron.NeuronWeights.Cast <object>().Take(784).ToArray();
hiddenNeurons.Rows.Add(row);
}
theDtFileOps.ConvertDataTableToTextFile(hiddenNeurons, "TheFinalNetworkHiddenWeights.txt");
}
public static void Main()
{
Console.WriteLine("Welcome to the Feed-forward backpropagation application!\n" +
"This application was written in college, but has been refactored to include this CLI.\n\n" +
"Would you like to perform data operations? This will create and train test data sets held locally.\n" +
"\nType 'Y' or hit enter to continue: ");
var performDataOperations = Console.ReadLine();
// Instantiate classes to utilize methods.
DataFileOperations theDtFileOps = new DataFileOperations();
RubricProblems theRubricProblems = new RubricProblems(theDtFileOps);
if (performDataOperations.ToLower().Equals("y"))
{
// Import complete data and create random train and test sets. Use 786 to have a column for bias at 785.
Console.WriteLine("\nData set training selected.\nConverting text from document to data table...");
theDtFileOps.ConvertTextToDataTable(786);
Console.WriteLine("\nData table created. \n\nGenerating train and test data sets...");
theDtFileOps.CreateTrainAndTestSets();
// Export Train and Test tables respectively.
Console.WriteLine("\nTrain and test data sets created. Writing complete table text file...");
theDtFileOps.ConvertDataTableToTextFile(theDtFileOps.CompleteTable, "CompleteTable.txt");
Console.WriteLine("\nComplete table file created. Writing train table text file...");
theDtFileOps.ConvertDataTableToTextFile(theDtFileOps.TrainTable, "TrainTable.txt");
Console.WriteLine("\nTrain table created. Writing test table text file...");
theDtFileOps.ConvertDataTableToTextFile(theDtFileOps.TestTable, "TestTable.txt");
Console.WriteLine("\nTest table file created.\n\nData file operations completed.");
}
Console.WriteLine("This application has been segmented into problems as required per course rubric.\n" +
"Please enter the problem number you would like to run (1-4) or enter to exit: ");
var problemValue = Console.ReadLine();
switch (int.Parse(problemValue))
{
case 1:
Console.WriteLine("\nProblem one selected. Starting processing...");
theRubricProblems.ProblemOne();
Console.WriteLine("\nProblem one completed.");
break;
case 2:
Console.WriteLine("\nProblem two selected. Starting processing...");
theRubricProblems.ProblemTwo();
Console.WriteLine("\nProblem two completed.");
break;
case 3:
Console.WriteLine("\nProblem three selected. Starting processing...");
theRubricProblems.ProblemThree();
Console.WriteLine("\nProblem three completed.");
break;
case 4:
Console.WriteLine("\nProblem four selected. Starting processing...");
theRubricProblems.ProblemFour();
Console.WriteLine("\nProblem four completed.");
break;
}
Console.WriteLine("\n\nExiting application...");
}