public void Train(double[][] trainData)
{
var particles = new PsoParticle[this.PsoProps.NumNetworks];
for (int i = 0; i < this.PsoProps.NumNetworks; i++)
{
particles[i] = new PsoParticle(new NeuralNetwork(this.NetworkProps, this.rng), this.PsoProps.ParticleProps, this.rng);
}
var bestAccuracy = 0.0;
for (int i = 0; i < this.PsoProps.Iterations; i++)
{
for (var p = 0; p < particles.Length; p++)
{
var accuracy = particles[p].UpdatePersonalBest(trainData);
if (accuracy > bestAccuracy)
{
this.Network = particles[p].Best.Clone();
bestAccuracy = accuracy;
}
if (accuracy > this.PsoProps.DesiredAccuracy)
{
return;
}
}
for (var p = 0; p < particles.Length; p++)
{
particles[p].MoveTowards(this.Network);
}
}
}
public void Train(double[][] trainData)
{
var particles = new PsoParticle[this.PsoProps.NumNetworks];
for (int i = 0; i < this.PsoProps.NumNetworks; i++)
{
particles[i] = new PsoParticle(new NeuralNetwork(this.NetworkProps, this.rng), this.PsoProps.ParticleProps, this.rng);
}
var bestAccuracy = 0.0;
for (int i = 0; i < this.PsoProps.Iterations; i++)
{
for (var p = 0; p < particles.Length; p++)
{
var accuracy = particles[p].UpdatePersonalBest(trainData);
if (accuracy > bestAccuracy)
{
this.Network = particles[p].Best.Clone();
bestAccuracy = accuracy;
}
if (accuracy > this.PsoProps.DesiredAccuracy)
{
return;
}
}
for (var p = 0; p < particles.Length; p++)
{
particles[p].MoveTowards(this.Network);
}
}
}
public void Train(double[][] trainData)
{
var particles = new PsoParticle[this.PsoProps.NumNetworks];
for (int i = 0; i < this.PsoProps.NumNetworks; i++)
{
particles[i] = new PsoParticle(new NeuralNetwork(this.NetworkProps, this.rng), this.PsoProps.ParticleProps, this.rng);
}
double pctComplete = this.PsoProps.Iterations / 10;
var bestAccuracy = 0.0;
for (int i = 0; i < this.PsoProps.Iterations; i++)
{
if ((i + 1) % pctComplete == 0)
Console.Write((100 * (i + 1) / this.PsoProps.Iterations).ToString() + @"%, ");
//if (((i+1) % 100) == 0)
// Console.Write((100 * (i+1) / this.PsoProps.Iterations).ToString() + @"%, ");
for (var p = 0; p < particles.Length; p++)
{
var accuracy = particles[p].UpdatePersonalBest(trainData);
if (accuracy > bestAccuracy)
{
this.Network = particles[p].Best.Clone();
bestAccuracy = accuracy;
}
if (accuracy > this.PsoProps.DesiredAccuracy)
{
Console.WriteLine((100 * (i + 1) / this.PsoProps.Iterations).ToString() + @"% (training ended)");
return;
}
}
for (var p = 0; p < particles.Length; p++)
{
particles[p].MoveTowards(this.Network);
}
}
}