public Network(int inputCount, LayerSettings[] hiddenLayerSettings, LayerSettings outputLayerSettings)
{
InputCount = inputCount;
if (hiddenLayerSettings is not null)
{
int layerInputCount = inputCount;
}
}
internal static float[] PrepareMemoryBuffer(int inputCount, LayerSettings[] hiddenLayerSettings, LayerSettings outputLayerSettings)
{
var totalNumber = 0;
totalNumber += inputCount; // First, keep buffer to write the network input
var layerInputCount = inputCount;
if (hiddenLayerSettings?.Length > 0)
{
var hiddenLayerNumber = 0;
foreach (var setting in hiddenLayerSettings)
{
if (setting.NeuronCount <= 0)
{
throw new ArgumentOutOfRangeException($"Every hidden layer must have non-negative neuron count: " +
$"HiddenLayerNumber {hiddenLayerNumber}; NeuronCount: {setting.NeuronCount}");
}
totalNumber += (layerInputCount + 1) * setting.NeuronCount; // For each neuron, we have 1 weight per input + 1 bias
totalNumber += setting.NeuronCount; // Outputs of the layer
layerInputCount = setting.NeuronCount; // One layer's output is nex one's input
hiddenLayerNumber++;
}
}
totalNumber += (layerInputCount + 1) * outputLayerSettings.NeuronCount;
totalNumber += outputLayerSettings.NeuronCount; // Network's outputs
return(new float[totalNumber]);
}