private void DrawLayerNeurons(bool fullState, NetworkDataModel model, LayerDataModel layer)
{
double threshold = model.Layers.First() == layer ? model.InputThreshold : 0;
Range.ForEach(layer.Neurons, neuron =>
{
if (fullState || (neuron.IsBias || neuron.Activation > threshold))
{
using (var brush = Tools.Draw.GetBrush(neuron.Activation))
{
if (neuron.IsBias)
{
G.FillEllipse(Brushes.Orange,
LayerX(model, layer) - BIAS_RADIUS,
VERTICAL_OFFSET + VerticalShift(model, layer) + neuron.Id * VerticalDistance(layer.Height) - BIAS_RADIUS,
BIAS_SIZE, BIAS_SIZE);
}
G.FillEllipse(brush,
LayerX(model, layer) - NEURON_RADIUS,
VERTICAL_OFFSET + VerticalShift(model, layer) + neuron.Id * VerticalDistance(layer.Height) - NEURON_RADIUS,
NEURON_SIZE, NEURON_SIZE);
}
}
});
}
public static void GaussHeEtAl(NetworkDataModel network, double?a)
{
// He initialization works better for layers with ReLu(s) activation.
if (!a.HasValue)
{
a = 1;
}
foreach (var layer in network.Layers)
{
foreach (var neuron in layer.Neurons)
{
foreach (var weight in neuron.Weights)
{
if (layer.Previous == null)
{
weight.Weight = Rand.GaussianRand.NextGaussian(0, a.Value);
}
else
{
weight.Weight = Rand.GaussianRand.NextGaussian(0, a.Value) * Math.Sqrt(2 / (double)layer.Previous.Neurons.Count);
}
}
}
}
}
private void Draw(bool fullState, NetworkDataModel model, bool isOnlyWeights, bool isOnlyChangedWeights, bool isHighlightChangedWeights)
{
if (model == null)
{
CtlPresenter.Clear();
return;
}
CtlPresenter.Clear();
//lock (Main.ApplyChangesLocker)
{
if (model.Layers.Count > 0)
{
foreach (var layer in model.Layers)
{
if (layer == model.Layers.Last())
{
break;
}
DrawLayersLinks(fullState, model, layer, layer.Next, isOnlyWeights, isOnlyChangedWeights, isHighlightChangedWeights);
}
foreach (var layer in model.Layers)
{
DrawLayerNeurons(fullState, model, layer);
}
}
}
CtlPresenter.Update();
}
public void SetInputDataAndDraw(NetworkDataModel model)
{
Threshold = model.InputThreshold;
var count = model.Layers[0].Neurons.Count(n => !n.IsBias);
if (Data == null || Data.Length != count)
{
Data = new double[count];
Stat = new double[Data.Length];
}
else
{
Array.Clear(Data, 0, Data.Length);
}
var neuron = model.Layers[0].Neurons[0];
while (neuron != null)
{
if (!neuron.IsBias)
{
Data[neuron.Id] = neuron.Activation;
}
neuron = neuron.Next;
}
Rearrange(PointsCount);
}
public void Do(NetworkDataModel model)
{
int bound = 10;
int number = Rand.Flat.Next(bound + 1);
if (number == 0)
{
return;
}
if (!_arrays.ContainsKey(bound))
{
_arrays.Add(bound, new double[bound]);
}
for (int i = 0; i < _arrays[bound].Length; ++i)
{
_arrays[bound][i] = i < number ? model.InputInitial1 : model.InputInitial0;
}
var shaffle = _arrays[bound].OrderBy(a => Rand.Flat.Next()).ToArray();
int k = 0;
Range.ForEach(model.Layers.First().Neurons.Where(n => !n.IsBias), n => n.Activation = shaffle[k++]);
}
public void SetInputDataAndDraw(NetworkDataModel model)
{
Threshold = model.InputThreshold;
Data = new double[model.Layers.First().Neurons.Where(n => !n.IsBias).Count()];
Range.ForEach(model.Layers.First().Neurons.Where(n => !n.IsBias), neuron => Data[neuron.Id] = neuron.Activation);
Rearrange(PointsCount);
}
public static void He_et_al(NetworkDataModel network, double?a)
{
if (!a.HasValue)
{
a = 1;
}
foreach (var layer in network.Layers)
{
foreach (var neuron in layer.Neurons)
{
foreach (var weight in neuron.Weights)
{
if (layer.Previous == null)
{
weight.Weight = Rand.GetFlatRandom();
}
else
{
weight.Weight = a.Value * Rand.GetFlatRandom() * Math.Sqrt(2 / layer.Previous.Neurons.Count);
}
}
}
}
}
public NetworkGPU(NetworkDataModel model)
{
var layers = GPU.Instance.Accelerator.Allocate <LayerGPU>(model.Layers.Count);
for (int i = 0; i < model.Layers.Count; ++i)
{
layers.CopyFrom(new LayerGPU(model.Layers[i]), i);
}
Layers = layers.View;
}
public double Do(NetworkDataModel model)
{
double s = 0;
var neuron = model.Layers.Last().Neurons[0];
while (neuron != null)
{
s += Math.Pow(neuron.Target - neuron.Activation, 2);
neuron = neuron.Next;
}
return(s);
}
public static void AbsGaussRandom(NetworkDataModel network, double?a)
{
foreach (var layer in network.Layers)
{
foreach (var neuron in layer.Neurons)
{
foreach (var weight in neuron.Weights)
{
weight.Weight = Math.Abs(Rand.GaussianRand.NextGaussian(0, a.HasValue ? a.Value : 1));
}
}
}
}
public static void FlatRandom(NetworkDataModel network, double?a)
{
foreach (var layer in network.Layers)
{
foreach (var neuron in layer.Neurons)
{
foreach (var weight in neuron.Weights)
{
weight.Weight = Rand.GetFlatRandom(a.HasValue ? a.Value : 1);
}
}
}
}
public void SetInputStat(NetworkDataModel model)
{
if (Stat != null)
{
int i = 0;
var neuron = model.Layers[0].Neurons[0];
while (neuron != null)
{
if (!neuron.IsBias)
{
Stat[i] += neuron.Activation > Threshold ? neuron.Activation : 0;
}
++i;
neuron = neuron.Next;
}
}
}
public void FeedForward(NetworkDataModel model)
{
for (int i = 0; i < model.Layers.Count - 1; ++i)
{
var neurons = model.Layers[i].GetNeurons();
var neuronsNext = model.Layers[i + 1].GetNeurons();
KernelProduct(neuronsNext.Height, neurons, neuronsNext);
Accelerator.Synchronize();
for (int n = 0; n < neuronsNext.Height; ++n)
{
neuronsNext.CopyTo(out double activation, new Index2(0, n));
model.Layers[i + 1].Neurons[n].Activation = model.Layers[i + 1].Neurons[n].ActivationFunction.Do(activation, model.Layers[i + 1].Neurons[n].ActivationFuncParamA);
neuronsNext.CopyFrom(model.Layers[i + 1].Neurons[n].Activation, new Index2(0, n));
}
}
}
public static void Centered(NetworkDataModel network, double?a)
{
if (!a.HasValue)
{
a = 1;
}
foreach (var layer in network.Layers)
{
foreach (var neuron in layer.Neurons)
{
foreach (var weight in neuron.Weights)
{
weight.Weight = -a.Value / 2 + a.Value * Rand.GetFlatRandom();
}
}
}
}
public static void Constant(NetworkDataModel network, double?a)
{
if (!a.HasValue)
{
a = 0;
}
foreach (var layer in network.Layers)
{
foreach (var neuron in layer.Neurons)
{
foreach (var weight in neuron.Weights)
{
weight.Weight = a.Value;
}
}
}
}
public static void WaveProgress(NetworkDataModel network, double?a)
{
if (!a.HasValue)
{
a = 6;
}
foreach (var layer in network.Layers)
{
foreach (var neuron in layer.Neurons)
{
foreach (var weight in neuron.Weights)
{
weight.Weight = InitializeMode.Centered(layer.Id + 1) * Math.Cos(weight.Id / Math.PI) * Math.Cos(neuron.Id / Math.PI);
}
}
}
}
private void DrawLayersLinks(bool fullState, NetworkDataModel model, LayerDataModel layer1, LayerDataModel layer2)
{
double threshold = model.Layers.First() == layer1 ? model.InputThreshold : 0;
Range.ForEach(layer1.Neurons, layer2.Neurons, (neuron1, neuron2) =>
{
if (!neuron2.IsBias || (neuron1.IsBias && neuron2.IsBiasConnected))
{
if (fullState || ((neuron1.IsBias || neuron1.Activation > threshold) && neuron1.AxW(neuron2) != 0))
{
using (var pen = Tools.Draw.GetPen(neuron1.AxW(neuron2), 1))
{
G.DrawLine(pen,
LayerX(model, layer1), VERTICAL_OFFSET + VerticalShift(model, layer1) + neuron1.Id * VerticalDistance(layer1.Height),
LayerX(model, layer2), VERTICAL_OFFSET + VerticalShift(model, layer2) + neuron2.Id * VerticalDistance(layer2.Height));
}
}
}
});
}
private void DrawLayerNeurons(bool fullState, NetworkDataModel model, LayerDataModel layer)
{
double threshold = model.Layers[0] == layer ? model.InputThreshold : 0;
var biasColor = Tools.Draw.GetPen(Colors.Orange);
NeuronDataModel prevNeuron = null;
foreach (var neuron in layer.Neurons)
{
if (fullState || neuron.IsBias || neuron.Activation > threshold || layer.Id > 0)
{
if (!Coordinator.ContainsKey(neuron))
{
Coordinator.Add(neuron, Points.Get(LayerX(model, layer), TOP_OFFSET + VerticalShift(model, layer) + neuron.Id * VerticalDistance(layer.Height)));
}
// Skip intersected neurons on first layer to improove performance.
if (!fullState && !neuron.IsBias && model.Layers[0] == layer && prevNeuron != null)
{
if (Coordinator[neuron].Y - Coordinator[prevNeuron].Y < NEURON_SIZE)
{
continue;
}
}
prevNeuron = neuron;
var pen = Tools.Draw.GetPen(neuron.Activation);
var brush = pen.Brush;
if (neuron.IsBias)
{
CtlPresenter.DrawEllipse(Brushes.Orange, biasColor, Coordinator[neuron], BIAS_RADIUS, BIAS_RADIUS);
}
CtlPresenter.DrawEllipse(brush, pen, Coordinator[neuron], NEURON_RADIUS, NEURON_RADIUS);
}
}
}
public void Draw(List <NetworkDataModel> models, NetworkDataModel selectedModel)
{
StartRender();
Clear();
DrawPlotter();
foreach (var model in models)
{
Vanish(model.DynamicStatistic.PercentData, GetPointPercentData);
Vanish(model.DynamicStatistic.CostData, GetPointCostData);
DrawData(model.DynamicStatistic.PercentData, model.Color, GetPointPercentData, false);
DrawData(model.DynamicStatistic.CostData, Color.FromArgb(80, model.Color), GetPointCostData, true);
}
if (selectedModel != null)
{
DrawLabel(selectedModel.DynamicStatistic.PercentData, selectedModel.Color);
}
CtlBox.Invalidate();
}
private void Draw(bool fullState, NetworkDataModel model)
{
StartRender();
Clear();
if (model == null)
{
CtlBox.Invalidate();
return;
}
lock (Main.ApplyChangesLocker)
{
if (model.Layers.Count > 0)
{
Range.ForEachTrimEnd(model.Layers, -1, layer => DrawLayersLinks(fullState, model, layer, layer.Next));
}
Range.ForEach(model.Layers, layer => DrawLayerNeurons(fullState, model, layer));
}
CtlBox.Invalidate();
}
public void Draw(ListX <NetworkDataModel> models, NetworkDataModel selectedModel)
{
if (IsBaseRedrawNeeded)
{
DrawPlotter();
IsBaseRedrawNeeded = false;
}
CtlPresenter.Clear();
if (models == null)
{
return;
}
var model = models[0];
while (model != null)
{
if (!model.IsEnabled)
{
continue;
}
DrawData(model.DynamicStatistics.CopyForRender.PercentData, Tools.Draw.GetColor(220, model.Color), GetPointPercentData, false);
DrawData(model.DynamicStatistics.CopyForRender.CostData, Tools.Draw.GetColor(150, model.Color), GetPointCostData, true);
model = model.Next;
}
if (selectedModel != null && selectedModel.DynamicStatistics.PercentData.Count > 0)
{
DrawLabel(selectedModel.DynamicStatistics.PercentData, selectedModel.Color);
}
CtlPresenter.Update();
}
public void Draw(List <NetworkDataModel> models, NetworkDataModel selectedModel)
{
int size = 9;
StartRender();
Clear();
long goodMax = 1;
long badMax = 1;
long axisOffset = 12;
long bound = 60;
var matrix = selectedModel.ErrorMatrix;
for (int y = 0; y < matrix.Output.Length; ++y)
{
for (int x = 0; x < matrix.Input.Length; ++x)
{
if (x == y)
{
goodMax = Math.Max(goodMax, matrix.Matrix[x, y]);
}
else
{
badMax = Math.Max(badMax, matrix.Matrix[x, y]);
}
}
}
for (int y = 0; y < matrix.Output.Length; ++y)
{
for (int x = 0; x < matrix.Input.Length; ++x)
{
var value = (double)matrix.Matrix[y, x] / (double)(x == y ? goodMax : badMax);
using (var brush = new SolidBrush(Tools.Draw.GetColorDradient(Color.LightGray, x == y ? Color.Green : Color.Red, 255, value)))
{
G.FillRectangle(brush, axisOffset + x * size, axisOffset + y * size, size, size);
G.DrawRectangle(Pens.Silver, axisOffset + x * size, axisOffset + y * size, size, size);
}
}
}
long outputMax = Math.Max(matrix.Output.Max(), 1);
for (int x = 0; x < matrix.Output.Length; ++x)
{
using (var brush = new SolidBrush(Tools.Draw.GetColorDradient(Color.White, matrix.Output[x] > matrix.Input[x] ? Color.Red : matrix.Output[x] < matrix.Input[x] ? Color.Blue : Color.Green, 100, (double)matrix.Output[x] / (double)outputMax)))
{
G.FillRectangle(brush, axisOffset + x * size, 2 + axisOffset + (matrix.Input.Length) * size, size, (int)(bound * (double)matrix.Output[x] / (double)outputMax));
G.DrawRectangle(Pens.Silver, axisOffset + x * size, 2 + axisOffset + (matrix.Input.Length) * size, size, (int)(bound * (double)matrix.Output[x] / (double)outputMax));
}
}
long inputMax = Math.Max(matrix.Input.Max(), 1);
for (int y = 0; y < matrix.Input.Length; ++y)
{
using (var brush = new SolidBrush(Tools.Draw.GetColorDradient(Color.White, Color.Green, 100, (double)matrix.Input[y] / (double)inputMax)))
{
G.FillRectangle(brush, 2 + axisOffset + (matrix.Output.Length) * size, axisOffset + y * size, (int)(bound * (double)matrix.Input[y] / (double)inputMax), size);
G.DrawRectangle(Pens.Silver, 2 + axisOffset + (matrix.Output.Length) * size, axisOffset + y * size, (int)(bound * (double)matrix.Input[y] / (double)inputMax), size);
}
}
G.DrawString("Output", Font, Brushes.Black, axisOffset, axisOffset - Font.Height - 1);
G.RotateTransform(-90);
G.DrawString("Input", Font, Brushes.Black, -axisOffset - (matrix.Output.Length) * size, axisOffset - Font.Height - 1);
G.RotateTransform(90);
G.Flush(System.Drawing.Drawing2D.FlushIntention.Sync);
CtlBox.Invalidate();
}
public NetworkDataModel CreateNetworkDataModel()
{
var model = new NetworkDataModel(Id, GetLayersSize())
{
Color = CtlColor.BackColor,
RandomizeMode = Randomizer,
RandomizerParamA = RandomizerParamA,
LearningRate = LearningRate,
InputInitial0 = ActivationFunction.Helper.GetInstance(InputLayer.ActivationFunc).Do(InputLayer.Initial0, null),
InputInitial1 = ActivationFunction.Helper.GetInstance(InputLayer.ActivationFunc).Do(InputLayer.Initial1, null)
};
model.Activate();
var layers = GetLayersControls();
for (int ln = 0; ln < layers.Count; ++ln)
{
model.Layers[ln].VisualId = layers[ln].Id;
var neurons = layers[ln].GetNeuronsControls();
for (int nn = 0; nn < neurons.Count; ++nn)
{
var neuronModel = model.Layers[ln].Neurons[nn];
neuronModel.VisualId = neurons[nn].Id;
neuronModel.IsBias = neurons[nn].IsBias;
neuronModel.IsBiasConnected = neurons[nn].IsBiasConnected;
neuronModel.ActivationFunction = ActivationFunction.Helper.GetInstance(neurons[nn].ActivationFunc);
neuronModel.ActivationDerivative = ActivationDerivative.Helper.GetInstance(neurons[nn].ActivationFunc);
neuronModel.ActivationFuncParamA = neurons[nn].ActivationFuncParamA;
neuronModel.WeightsInitializer = neurons[nn].WeightsInitializer;
neuronModel.WeightsInitializerParamA = neurons[nn].WeightsInitializerParamA;
double initValue = InitializeMode.Helper.Invoke(neurons[nn].WeightsInitializer, neurons[nn].WeightsInitializerParamA);
if (!InitializeMode.Helper.IsSkipValue(initValue))
{
foreach (var weight in neuronModel.Weights)
{
weight.Weight = InitializeMode.Helper.Invoke(neurons[nn].WeightsInitializer, neurons[nn].WeightsInitializerParamA);
}
}
if (neuronModel.IsBias)
{
neuronModel.ActivationInitializer = neurons[nn].ActivationInitializer;
neuronModel.ActivationInitializerParamA = neurons[nn].ActivationInitializerParamA;
initValue = InitializeMode.Helper.Invoke(neurons[nn].ActivationInitializer, neurons[nn].ActivationInitializerParamA);
if (!InitializeMode.Helper.IsSkipValue(initValue))
{
neuronModel.Activation = initValue;
}
}
}
}
model.Layers.Last().VisualId = Const.OutputLayerId;
{
var neurons = OutputLayer.GetNeuronsControls();
for (int i = 0; i < neurons.Count; ++i)
{
model.Layers.Last().Neurons[i].VisualId = neurons[i].Id;
}
}
return(model);
}
public NetworkDataModel CreateNetworkDataModel(INetworkTask task, bool isCopy)
{
ErrorMatrix em1 = null;
if (task != null)
{
em1 = new ErrorMatrix(task.GetClasses());
var em2 = new ErrorMatrix(task.GetClasses());
em1.Next = em2;
em2.Next = em1;
}
var model = new NetworkDataModel(Id, GetLayersSize())
{
ErrorMatrix = em1,
Classes = task?.GetClasses(),
IsEnabled = CtlIsNetworkEnabled.IsOn,
Color = CtlColor.Foreground.GetColor(),
RandomizeMode = RandomizeMode,
RandomizerParamA = RandomizerParamA,
LearningRate = LearningRate,
InputInitial0 = ActivationFunction.Helper.GetInstance(InputLayer.ActivationFunc).Do(InputLayer.Initial0, InputLayer.ActivationFuncParamA),
InputInitial1 = ActivationFunction.Helper.GetInstance(InputLayer.ActivationFunc).Do(InputLayer.Initial1, InputLayer.ActivationFuncParamA),
CostFunction = CostFunction.Helper.GetInstance(CtlCostFunction.SelectedValue.ToString()),
IsAdjustFirstLayerWeights = InputLayer.IsAdjustFirstLayerWeights
};
model.ActivateNetwork();
LayerDataModel prevLayer = null;
var layers = GetLayersControls();
for (int ln = 0; ln < layers.Count; ++ln)
{
if (ln > 0)
{
prevLayer = model.Layers[ln - 1];
}
model.Layers[ln].VisualId = layers[ln].Id;
var neurons = layers[ln].GetNeuronsControls();
for (int nn = 0; nn < neurons.Count; ++nn)
{
var neuronModel = model.Layers[ln].Neurons[nn];
neuronModel.VisualId = neurons[nn].Id;
neuronModel.IsBias = neurons[nn].IsBias;
neuronModel.IsBiasConnected = neurons[nn].IsBiasConnected;
neuronModel.ActivationFunction = ActivationFunction.Helper.GetInstance(neurons[nn].ActivationFunc);
neuronModel.ActivationFuncParamA = neurons[nn].ActivationFuncParamA;
if (ln == 0 && !neuronModel.IsBias)
{
neuronModel.WeightsInitializer = InputLayer.WeightsInitializer;
neuronModel.WeightsInitializerParamA = InputLayer.WeightsInitializerParamA;
double initValue = InitializeMode.Helper.Invoke(neuronModel.WeightsInitializer, neuronModel.WeightsInitializerParamA);
if (!InitializeMode.Helper.IsSkipValue(initValue))
{
neuronModel.Weights.ForEach(w => w.Weight = InitializeMode.Helper.Invoke(neuronModel.WeightsInitializer, neuronModel.WeightsInitializerParamA));
}
}
else
{
neuronModel.WeightsInitializer = neurons[nn].WeightsInitializer;
neuronModel.WeightsInitializerParamA = neurons[nn].WeightsInitializerParamA;
double initValue = InitializeMode.Helper.Invoke(neuronModel.WeightsInitializer, neuronModel.WeightsInitializerParamA);
if (!InitializeMode.Helper.IsSkipValue(initValue))
{
neuronModel.Weights.ForEach(w => w.Weight = InitializeMode.Helper.Invoke(neuronModel.WeightsInitializer, neuronModel.WeightsInitializerParamA));
}
}
if (neuronModel.IsBias)
{
neuronModel.ActivationInitializer = neurons[nn].ActivationInitializer;
neuronModel.ActivationInitializerParamA = neurons[nn].ActivationInitializerParamA;
double initValue = InitializeMode.Helper.Invoke(neurons[nn].ActivationInitializer, neurons[nn].ActivationInitializerParamA);
if (!InitializeMode.Helper.IsSkipValue(initValue))
{
neuronModel.Activation = initValue;
}
}
if (!isCopy && prevLayer != null && prevLayer.Height > 0)
{
neuronModel.ForwardHelper = new ListX <ForwardNeuron>(prevLayer.Height);
var prevNeuron = prevLayer.Neurons[0];
while (prevNeuron != null)
{
if (!neuronModel.IsBias || (neuronModel.IsBiasConnected && prevNeuron.IsBias))
{
neuronModel.ForwardHelper.Add(new ForwardNeuron(prevNeuron, prevNeuron.WeightTo(neuronModel)));
}
prevNeuron = prevNeuron.Next;
}
}
}
}
model.Layers.Last().VisualId = Const.OutputLayerId;
{
var neurons = OutputLayer.GetNeuronsControls();
for (int i = 0; i < neurons.Count; ++i)
{
model.Layers.Last().Neurons[i].VisualId = neurons[i].Id;
}
}
if (!isCopy)
{
model.Copy = CreateNetworkDataModel(task, true);
}
return(model);
}
private void DrawLayersLinks(bool fullState, NetworkDataModel model, LayerDataModel layer1, LayerDataModel layer2, bool isOnlyWeights, bool isOnlyChangedWeights, bool isHighlightChangedWeights)
{
double threshold = model.Layers[0] == layer1 ? model.InputThreshold : 0;
NeuronDataModel prevNeuron = null;
foreach (var neuron1 in layer1.Neurons)
{
if (!Coordinator.ContainsKey(neuron1))
{
Coordinator.Add(neuron1, Points.Get(LayerX(model, layer1), TOP_OFFSET + VerticalShift(model, layer1) + neuron1.Id * VerticalDistance(layer1.Height)));
}
// Skip intersected neurons on first layer to improove performance.
if (!fullState && !neuron1.IsBias && model.Layers[0] == layer1 && prevNeuron != null)
{
if (Coordinator[neuron1].Y - Coordinator[prevNeuron].Y < NEURON_SIZE)
{
continue;
}
}
if (fullState || neuron1.IsBias || neuron1.Activation > threshold)
{
foreach (var neuron2 in layer2.Neurons)
{
if (!neuron2.IsBias || (neuron2.IsBiasConnected && neuron1.IsBias))
{
if (fullState || ((neuron1.IsBias || neuron1.Activation > threshold) && neuron1.AxW(neuron2) != 0))
{
Pen pen = null;
Pen penChange = null;
bool isWeightChanged = false;
var weight = neuron1.WeightTo(neuron2);
if (WeightsData.TryGetValue(weight, out double prevWeight))
{
if (prevWeight != weight.Weight)
{
isWeightChanged = true;
WeightsData[weight] = weight.Weight;
}
}
else
{
prevWeight = 0;
isWeightChanged = true;
WeightsData.Add(weight, weight.Weight);
}
double fraction = Math.Min(1, Math.Abs((prevWeight - weight.Weight) / prevWeight));
if (fraction <= 0.001)
{
isWeightChanged = false;
}
if (isWeightChanged && isHighlightChangedWeights)
{
penChange = Tools.Draw.GetPen(Colors.Lime);
}
if (isOnlyWeights)
{
if ((isWeightChanged && isOnlyChangedWeights) || !isOnlyChangedWeights)
{
pen = Tools.Draw.GetPen(weight.Weight, 1);
}
}
else
{
pen = Tools.Draw.GetPen(neuron1.AxW(neuron2), 1);
}
if (!isOnlyChangedWeights && isHighlightChangedWeights && isWeightChanged)
{
pen = penChange;
penChange = null;
}
if (pen != null)
{
if (!Coordinator.ContainsKey(neuron2))
{
Coordinator.Add(neuron2, Points.Get(LayerX(model, layer2), TOP_OFFSET + VerticalShift(model, layer2) + neuron2.Id * VerticalDistance(layer2.Height)));
}
CtlPresenter.DrawLine(pen, Coordinator[neuron1], Coordinator[neuron2]);
prevNeuron = neuron1;
}
if (penChange != null)
{
CtlPresenter.DrawLine(penChange, Coordinator[neuron1], Coordinator[neuron2]);
prevNeuron = neuron1;
}
}
}
}
}
}
}
private float VerticalShift(NetworkDataModel model, LayerDataModel layer)
{
return((MaxHeight(model) - layer.Height * VerticalDistance(layer.Height)) / 2);
}
private int LayerX(NetworkDataModel model, LayerDataModel layer)
{
return(HORIZONTAL_OFFSET + LayerDistance(model) * layer.Id);
}
public int LayerDistance(NetworkDataModel model)
{
return((int)(ActualWidth - 2 * HORIZONTAL_OFFSET) / (model.Layers.Count - 1));
}
public void Do(NetworkDataModel model)
{
Range.For(Rand.Flat.Next(11), i => model.Layers.First().Neurons.RandomElementTrimEnd(model.Layers.First().BiasCount).Activation = model.InputInitial1);
}
private new float MaxHeight(NetworkDataModel model)
{
return(model.Layers.Max(layer => layer.Height * VerticalDistance(layer.Height)));
}
