/// <summary>
/// Applies delta sums, modifying coefficients and biases of the layer as a result of back propogation.
/// </summary>
/// <param name="showDetails">Set to <i>true</i> - prints delta sums using PrintClass, set to <i>false</i> - no printing.</param>
/// <param name="batchSize">The number of examples of current teaching for L2 regularization realization.</param>
public void ApplyDeltaSums(bool showDetails, int batchSize)
{
// GOING THROUGH ALL NEURONS...
for (int neuronInd = 0; neuronInd < neuronDeltas.Length; neuronInd++)
{
// CHANGING BIASES
if (showDetails)
{
PrintClass.Print("\n - Neuron " + neuronInd.ToString() + " bias delta sum: " + biasDeltaSums[neuronInd].ToString() + "\n coeffs delta sums: ");
}
biases[neuronInd] -= biasDeltaSums[neuronInd];
biasDeltaSums[neuronInd] = 0;
neuronDeltas[neuronInd] = 0;
// GOING THROUGH ALL COEFFS OF THE NEURON...
for (int coeffInd = 0; coeffInd < coeffs[neuronInd].Length; coeffInd++)
{
// CHANGING COEFFS APPLYING L2 REGULARIZATION
if (showDetails)
{
PrintClass.Print(coeffDeltaSums[neuronInd][coeffInd].ToString() + " ");
}
coeffs[neuronInd][coeffInd] = (1 - ANNetwork.SayNetworkLyambda() * ANNetwork.SayNetworkLearningSpeed() / batchSize) * coeffs[neuronInd][coeffInd] - coeffDeltaSums[neuronInd][coeffInd];
coeffDeltaSums[neuronInd][coeffInd] = 0;
}
}
if (showDetails)
{
Console.Write("\n");
}
}
// LISTEX COMMAND
public void ListExs()
{
string result = examples.ShowFileList();
if (result.Length > 0)
{
PrintClass.PrintLine(result);
}
}
// CLEAR EXAMPLES
public string ClearExamples()
{
Examples tmpExamples = Examples.Init();
if (tmpExamples.SayExamplesNum() == 0)
{
return("\n-> Error clearing examples - no examples exist.");
}
tmpExamples.ClearExamples();
PrintClass.PrintLine("\n-> Training and testing examples lists cleared - no examples exist any more.");
return("");
}
/// <summary>
/// Prints layer parameters (coefficients and biases) using PrintClass
/// </summary>
public void ShowParams(bool showDetails)
{
PrintClass.PrintLine(" Neurons number: " + neuronDeltas.Length);
PrintClass.PrintLine(" Inputs number: " + coeffs[0].Length);
PrintClass.PrintLine(" Activation function type: " + AFType.ToString().Remove(AFType.ToString().IndexOf("AFType")));
if (showDetails)
{
string tmpString;
for (int i = 0; i < neuronDeltas.Length; i++)
{
PrintClass.PrintLine("\n - Neuron " + i.ToString() + ": bias " + biases[i] + ", coefficients :");
tmpString = " ";
for (int j = 0; j < coeffs[i].Length; j++)
{
tmpString += coeffs[i][j].ToString() + " ";
}
PrintClass.PrintLine(tmpString);
}
}
}
// COST EXAMPLES
public string Cost()
{
if (network == null)
{
return("\n-> Error calculating cost - no network exists.");
}
if (!network.examples.ExamplesExist())
{
return("\n-> Error calculating cost - no examples exists.");
}
string result = network.CalcAllCost(out float cost, true);
if (result.Length > 0)
{
return(result);
}
PrintClass.PrintLine("\n-> Total cost for existing " + network.examples.SayExamplesNum().ToString() + " examples is: " + cost.ToString() + ".");
return("");
}
// HELP MESSAGE
public void Help()
{
PrintClass.PrintLine(helpMessage);
}
// RUN EXAMPLE
public string RunExample(string tmpString)
{
tmpString = tmpString.Trim();
// CHECKING EXAMPLES, NETWORK AND INPUT
if (network == null)
{
return("\n-> Error running an example - the network doesn't exist.");
}
if (network.examples.SayExamplesNum() == 0)
{
return("\n-> Error running an example - no existing examples.");
}
if (tmpString.Length == 0)
{
return("\n-> Error running an example - not enough input data.");
}
// PARSING EXAMPLE INDEX
int exampleInd = -1;
bool isLast = true;
string result = Parsing.ParseInt(ref tmpString, ref exampleInd, ref isLast, "examplesInd", Parsing.AfterParsingCheck.NonNegative);
if (result.Length > 0)
{
return(result);
}
bool allLayersOutputs = false;
// PARSING OPTION OF PRINTING ALL LAYERS OUTPUTS
if (!isLast)
{
int allLOutputs = -1;
result = Parsing.ParseInt(ref tmpString, ref allLOutputs, ref isLast, "printAllLayersOutputs", Parsing.AfterParsingCheck.NonNegative);
if (result.Length > 0)
{
return(result);
}
allLayersOutputs = allLOutputs == 0 ? false : true;
}
float[][] outputs = null;
result = network.examples.GiveExampleLink(exampleInd, out Example theExample);
if (result.Length < 0)
{
return(result);
}
result = network.React(theExample.SayX(), ref outputs, out int networkAnswer);
if (result.Length > 0)
{
return(result);
}
PrintClass.Print("\n\n-> On example N " + exampleInd.ToString() + "\n(" + theExample.SayX(0));
for (int inputInd = 1; inputInd < theExample.SayXLength(); inputInd++)
{
PrintClass.Print(", " + theExample.SayX(inputInd));
}
PrintClass.Print(")\n network replied: " + networkAnswer.ToString() + " (" + outputs[outputs.Length - 1][0].ToString());
for (int outputInd = 1; outputInd < outputs[outputs.Length - 1].Length; outputInd++)
{
PrintClass.Print(", " + outputs[outputs.Length - 1][outputInd].ToString());
}
PrintClass.Print(").\n while expected is: " + theExample.SayExpected().ToString() + " (" + theExample.SayY(0).ToString());
for (int outputInd = 1; outputInd < theExample.SayYLength(); outputInd++)
{
PrintClass.Print(", " + theExample.SayY(outputInd).ToString());
}
PrintClass.Print(").\n");
if (allLayersOutputs)
{
for (int layerInd = 0; layerInd < outputs.Length; layerInd++)
{
PrintClass.Print("\n - Layer " + layerInd.ToString() + " outputs:\n" + outputs[layerInd][0].ToString());
for (int outputInd = 1; outputInd < outputs[layerInd].Length; outputInd++)
{
PrintClass.Print(", " + outputs[layerInd][outputInd].ToString());
}
}
}
return("");
}
