public double[] Predict(Core.fData image)
{
if (Input.Output.Count == 3)
{
ColorImage inputlayer = (ColorImage)Input;
inputlayer.fData = image;
}
else
{
GrayImage inputlayer = (GrayImage)Input;
inputlayer.fData = image;
}
Next(ANN.Enums.Propagate.Signal);
Foundation.Node node;
double[] probs = new double[Output.Output[0].Rows];
for (int l = 0; l < Output.Output[0].Rows; l++)
{
node = Output.Output[0].GetElement(l, 0);
probs[l] = (double)((double?[])node.Element)[Global.Sig];
}
return(probs);
}
public override string Next()
{
// 0. validate parameters
if ((learningRate == null) || (momentum == null))
{
throw new Exception();
}
// 3. configure randomizer...
int[] batchRandomizer = new int[batchImages.Count()];
int[] dataSetRandomizer = new int[nofSet];
log += "\nTraining data...";
for (int i = 0; i < Epochs; i++)
{
//log += "\n\n////////// Epoch[" + i + "]... ///////////////////////////////////////////////////////";
Global.NextIntArray(0, nofSet, dataSetRandomizer);
Global.NextIntArray(0, batchImages.Count(), batchRandomizer);
// single epoch
for (int j = 0; j < batchImages.Count; j++) //batch
{
//DisplayWeights();
batchLoss = 0;
for (int k = 0; k < nofSet; k++)
{
if (Model.Input.Output.Count == 3)
{
ColorImage inputlayer = (ColorImage)Model.Input;
inputlayer.fData = batchImages[batchRandomizer[j]].fData[dataSetRandomizer[k]];
}
else
{
GrayImage inputlayer = (GrayImage)Model.Input;
inputlayer.fData = batchImages[batchRandomizer[j]].fData[dataSetRandomizer[k]];
}
//log += "\n\nModel: " + Model.ToString();
//DisplayWeights();
// 4.1 propagate signal
Model.Next(Propagate.Signal);
//log += "\n\nModel: " + Model.ToString();
// 4.2 set targets
Layer lyr = Model.Output;
Node node;
probs = new double[lyr.Output[0].Rows];
for (int l = 0; l < batchImages[batchRandomizer[j]].fData[dataSetRandomizer[k]].Label.Length; l++)
{
node = lyr.Output[0].GetElement(l, 0);
probs[l] = (double)((double?[])node.Element)[Global.Sig];
((double?[])node.Element)[Global.Err] = batchImages[batchRandomizer[j]].fData[dataSetRandomizer[k]].Label[l];
}
// Set softmax output
double[] softmaxOutput = Softmax(probs);
for (int l = 0; l < probs.Length; l++)
{
node = lyr.Output[0].GetElement(l, 0);
((double?[])node.Element)[Global.Sig] = softmaxOutput[l];
}
singleLoss = lossfunc.GetLoss(softmaxOutput, batchImages[batchRandomizer[j]].fData[dataSetRandomizer[k]].Label);
// Calculate batch loss
batchLoss += singleLoss;
//4.4 propagate error and set new weights
Model.Next(Propagate.Error);
SaveError();
//log += "\n" + Model.ToString();
}
// 4.5 adjust weights
AdjustWeights();
//DisplayWeights();
ClearError();
batchLoss = batchLoss / (double)batchSize;
//log += "\n" + Model.ToString();
Console.WriteLine("Epoch[" + i.ToString() + "]" + "Batch[" + j.ToString() + "]" + "Loss: " + batchLoss.ToString("e4"));
log += "\n\nEpoch[" + i.ToString() + "]" + "Batch[" + j.ToString() + "]" + "Loss: " + batchLoss.ToString("e4");
}
}
return(log);
}
public override string Next()
{
// 0. validate parameters
if ((learningRate == null) || (momentum == null))
{
throw new Exception();
}
// 3. configure randomizer...
int[] dataSetRandomizer = new int[nofSet];
log += "\nTraining data...";
// 4. train epoch by epoch...
for (int i = 0; i < Epochs; i++)
{
Console.WriteLine("Start");
log += "\n\n////////// Epoch[" + i + "]... ///////////////////////////////////////////////////////";
Global.NextIntArray(0, nofSet, dataSetRandomizer);
//log += "\n\nDataSet randomizer sequence: ";
//for (int s = 0; s < dataSetRandomizer.Length; s++)
// log += dataSetRandomizer[s] + ", ";
//log = log.TrimEnd(new char[] { ',', ' ' });
// single epoch
double mean_loss = 0;
for (int j = 0; j < nofSet; j++)
{
if (Model.Input.Output.Count == 3)
{
ColorImage inputlayer = (ColorImage)Model.Input;
inputlayer.fData = DataSet.fData[dataSetRandomizer[j]];
}
else
{
GrayImage inputlayer = (GrayImage)Model.Input;
inputlayer.fData = DataSet.fData[dataSetRandomizer[j]];
}
//log += "\n\nModel: " + Model.ToString();
// 4.1 propagate signal
//DisplayWeights();
Model.Next(Propagate.Signal);
//log += "\n\nModel: " + Model.ToString();
// 4.2 set targets
Layer lyr = Model.Output;
Node node;
probs = new double[lyr.Output[0].Rows];
for (int l = 0; l < DataSet.fData[dataSetRandomizer[j]].Label.Length; l++)
{
node = lyr.Output[0].GetElement(l, 0);
probs[l] = (double)((double?[])node.Element)[Global.Sig];
((double?[])node.Element)[Global.Err] = DataSet.fData[dataSetRandomizer[j]].Label[l];
}
//Set softmax output
double[] softmaxOutput = Softmax(probs);
for (int l = 0; l < probs.Length; l++)
{
node = lyr.Output[0].GetElement(l, 0);
((double?[])node.Element)[Global.Sig] = softmaxOutput[l];
}
// 4.4 propagate error and set new weights
Model.Next(Propagate.Error);
// 4.5 adjust weights
AdjustWeights();
//4.6 Calculate loss
loss = lossfunc.GetLoss(softmaxOutput, DataSet.fData[dataSetRandomizer[j]].Label);
mean_loss += loss;
//log += "\n\nEpoch[" + i.ToString() + "]" + "Image[" + j.ToString() + "]" + "Loss: " + loss.ToString("e4");
Console.WriteLine("Epoch[" + i.ToString() + "]" + "Batch[" + j.ToString() + "]" + "Loss: " + loss.ToString("e4"));
//log += "\n" + Model.ToString();
}
Console.WriteLine("Epoch[" + i.ToString() + "]" + "Loss: " + (mean_loss / nofSet).ToString("e4"));
}
return(log);
}
