public GFNN1LayerSystem(IIntegrator integrator, double middleFrequency = 440.0, int octaves = 4, int nodesPerOctave = 120, ToneNode soundSource = null)
: base(integrator)
{
// "A dynamical systems approach to musical tonality.". Large 2010. Section 4 "Predicting Tonality".
// Generate list of natural frequencies. We step
// linearly through the log-frequency space.
int numOctaves = octaves;
int numFrequenciesPerOctave = nodesPerOctave;
SortedSet <double> frequencies = new SortedSet <double>();
double xMiddle = Math.Log(middleFrequency);
double numOctavesHalfPow2 = Math.Pow(2.0, numOctaves / 2.0);
double xLowerBound = Math.Log(middleFrequency / numOctavesHalfPow2);
double xUpperBound = Math.Log(middleFrequency * numOctavesHalfPow2);
double xStepSize = (xUpperBound - xLowerBound) / (numOctaves * numFrequenciesPerOctave);
// Some variables.
double x;
// Some methods
Func <double, double> logFrequencyToFrequency = (lf) => Math.Exp(lf);
// Add midpoint.
x = xMiddle;
frequencies.Add(logFrequencyToFrequency(x));
// Add lower frequencies.
while (x >= xLowerBound)
{
x -= xStepSize;
frequencies.Add(logFrequencyToFrequency(x));
}
// Add upper frequencies.
x = xMiddle;
while (x <= xUpperBound)
{
x += xStepSize;
frequencies.Add(logFrequencyToFrequency(x));
}
// Create nodes.
double alpha;
double beta;
double delta;
double epsilon;
double omega;
// Configuration 1
{
// "A canonical model for gradient frequency neural networks.
// Equation 20 test parameters.
alpha = 0.0;
beta = -10.0;
delta = -9.0;
epsilon = 0.3;
omega = 2.0 * Math.PI;
}
List <DynamicalNode> nodes = new List <DynamicalNode>();
foreach (double frequency in frequencies)
{
NeuralOscillatorNode node = new NeuralOscillatorNode(
0.0,
frequency,
epsilon,
alpha,
beta,
beta,
delta,
omega);
node.Name = string.Format("ω{0:F3}", frequency);
node.HistorySize = 2;
nodes.Add(node);
}
// Add audio source node if present.
if (soundSource != null)
{
foreach (var node in nodes)
{
node.AddIncomingNode(soundSource, 1.0, 0.0, 0.0);
}
nodes.Add(soundSource);
this.SoundSource = soundSource;
}
// Store nodes.
this.Nodes = nodes;
}
protected List <NeuralOscillatorNode> GenerateLayer(
string namePrefix,
double middleFrequency,
int numOctaves,
int numFrequenciesPerOctave,
double alpha,
double beta1,
double beta2,
double delta,
double epsilon,
double omega)
{
SortedSet <double> frequencies = new SortedSet <double>();
double xMiddle = Math.Log(middleFrequency);
double numOctavesHalfPow2 = Math.Pow(2.0, numOctaves / 2.0);
double xLowerBound = Math.Log(middleFrequency / numOctavesHalfPow2);
double xUpperBound = Math.Log(middleFrequency * numOctavesHalfPow2);
double xStepSize = (xUpperBound - xLowerBound) / (numOctaves * numFrequenciesPerOctave);
// Some variables.
double x;
// Some methods
Func <double, double> logFrequencyToFrequency = (lf) => Math.Exp(lf);
// Add midpoint.
x = xMiddle;
frequencies.Add(logFrequencyToFrequency(x));
// Add lower frequencies.
while (x >= xLowerBound)
{
x -= xStepSize;
frequencies.Add(logFrequencyToFrequency(x));
}
// Add upper frequencies.
x = xMiddle;
while (x <= xUpperBound)
{
x += xStepSize;
frequencies.Add(logFrequencyToFrequency(x));
}
List <NeuralOscillatorNode> nodes = new List <NeuralOscillatorNode>();
foreach (double frequency in frequencies)
{
NeuralOscillatorNode node = new NeuralOscillatorNode(
0.0,
frequency,
epsilon,
alpha,
beta1,
beta2,
delta,
omega);
node.Name = string.Format("{1}:ω{0:F3}", frequency, namePrefix);
node.HistorySize = 2;
nodes.Add(node);
}
return(nodes);
}
