public Configuration(Configuration template)
{
model = template.model;
SIAs = new List<Tuple<double, double>>();
SIAs.AddRange(template.SIAs);
Weights = new List<double>();
Weights.AddRange(template.Weights);
}
public List<Configuration> Run(List<Configuration> elements)
{
//Runs ONE iteration
// copy default configuration
List<Configuration> population = new List<Configuration>();
population.AddRange(elements);
Configuration leader = population[0];
Configuration worst = population[0];
foreach (Configuration c in population)
{
if (c.GE < leader.GE)
leader = c;
if (c.GE > worst.GE)
worst = c;
}
if (worst.GE - leader.GE < ps["MinDiv"]) //everyone too close
return population;
List<Configuration> ng = new List<Configuration>();
for (int k = 0; k < population.Count; k++)
{
Configuration c = population[k];
// Dealing with leader, put him in the candidate list
if (c == leader)
{
ng.Add(c);
continue;
}
List<Configuration> jumps = new List<Configuration>();
for (int i = 0; i < ps["PathLength"] / ps["Step"]; i++)
{
int prtsize = leader.SIAs.Count * 3 + leader.Weights.Count;
List<int> prtv = GetPRTVector(prtsize);
Configuration leaderdiff = leader.GetDifference(c);
Configuration onejump = new Configuration(c);
c.AddVector(leaderdiff, i, ps["Step"], prtv);
jumps.Add(onejump);
}
int bestindex = -1;
for (int j = 0; j < jumps.Count; j++)
{
if (jumps[j].GE < c.GE)
bestindex = j;
}
ng.Add((bestindex == -1) ? population[k] : jumps[bestindex]);
}
return population;
}
public void UpdateConfiguration(Configuration c)
{
// only weights and slopes will be altered
//assemble list of all neurons and then all synapses
List<Neuron> allneurons = new List<Neuron>();
List<Synapse> allsynapses = new List<Synapse>();
foreach (List<Neuron> nlist in Neurons.Values)
allneurons.AddRange(nlist);
foreach (List<Synapse> slist in Synapses.Values)
allsynapses.AddRange(slist);
// update data
if (c.SIAs.Count != allneurons.Count || c.Weights.Count != allsynapses.Count)
return;
for (int i = 0; i < allneurons.Count; i++)
{
allneurons[i].Slope = c.SIAs[i].Item1;
allneurons[i].Augment = c.SIAs[i].Item2;
}
for (int i = 0; i < allsynapses.Count; i++)
{
allsynapses[i].Weight = c.Weights[i];
}
}
public void AddVector(Configuration leaderdiff, int step, double stepsize, List<int> prt)
{
int prtcount = 0;
for (int i = 0; i < SIAs.Count; i++)
{
SIAs[i] = new Tuple<double, double>
(
SIAs[i].Item1 + leaderdiff.SIAs[i].Item1 * step * stepsize * prt[prtcount],
SIAs[i].Item2 + leaderdiff.SIAs[i].Item2 * step * stepsize * prt[prtcount + 1]
);
prtcount += 2; // must match tuple size!!!
}
for (int i = 0; i < Weights.Count; i++)
{
Weights[i] += leaderdiff.Weights[i] * step * stepsize * prt[prtcount];
prtcount++;
}
}
public Configuration GetDifference(Configuration c)
{
Configuration result = new Configuration(this);
for (int i = 0; i < this.SIAs.Count; i++)
result.SIAs[i] = new Tuple<double, double>
(result.SIAs[i].Item1 - c.SIAs[i].Item1, result.SIAs[i].Item2 - c.SIAs[i].Item2);
for (int i = 0; i < this.Weights.Count; i++)
result.Weights[i] -= c.Weights[i];
return result;
}