/// <summary>
/// We keep the population bound to a probablity of fitness
/// </summary>
/// <param name="pop">
/// </param>
/// <example>
/// <![CDATA[
/// The population is represented as a coupled-pair of weighted fitness probablity
/// to the the State
///
/// Weighted Prob State
/// 1 s_1
/// 0.7961956521 s_2
/// 0.5448369565 s_3
/// ... ...
/// 0.1698369565 s_i-1
/// 0.0543478260 s_i
///
/// ]]>
/// </example>
private void CalcWtFitness2State(IList<IState> pop)
{
pop.Shuffle();
var calc = new List<Tuple<double, IState>>();
var total = pop.Select(x => FitnessFunction.Eval(x)).Sum();
var runningTotal = 0.0D;
foreach (var s in pop)
{
runningTotal += FitnessFunction.Eval(s) / total;
calc.Add(new Tuple<double, IState>(runningTotal, s));
}
_wtFitness2State = calc.OrderByDescending(x => x.Item1).ToList();
}
/// <summary>
/// Does NOT use <see cref="IProblem.IsGoalTest"/> to return an <see cref="IState"/> but simply gets
/// the <see cref="IState"/> which max'ed the <see cref="FitnessFunction"/> after <see cref="NumberOfGenerations"/>
/// generations.
/// </summary>
/// <returns></returns>
public IState Search()
{
var popLen = _wtFitness2State.Count;
//go through the number of supplied generations
for (var i = 0; i < _numberOfGenerations; i++)
{
//for each generation the population is replaced by children of more fit offspring
var nextGeneration = new List<IState>();
for (var j = 0; j < popLen; j++)
{
var ps = PickFolks();
var child = _mutationFx(_reproductionFx(ps.Item1, ps.Item2));
nextGeneration.Add(child);
}
CalcWtFitness2State(nextGeneration);
}
//there can be only one.
return _wtFitness2State.Select(x => x.Item2).OrderBy(x => FitnessFunction.Eval(x)).First();
}
