public void Test1()
{
var dicsOfProbs = new Dictionary <string, double>
{
{ "A", 0.1 },
{ "B", 0.2 },
{ "C", 0.4 },
{ "D", 0.2 },
{ "E", 0.1 },
};
var probabilityRandom = new ProbabilityRandom <string>();
foreach (var item in dicsOfProbs)
{
probabilityRandom.SetProb(item.Key, item.Value);
}
probabilityRandom.SetProb("D", 0.1);
probabilityRandom.SetProb("F", 0.1);
var rand = new Random();
var actualProbs = new Dictionary <string, double>
{
{ "A", 0 },
{ "B", 0 },
{ "C", 0 },
{ "D", 0 },
{ "E", 0 },
{ "F", 0 },
};
int maxRand = 50000;
for (int i = 0; i < maxRand; i++)
{
var num = probabilityRandom.Next();
actualProbs[num] = actualProbs[num] + 1;
}
double sum = 0;
foreach (var item in actualProbs)
{
console.WriteLine($"P({item.Key}) = {item.Value / maxRand} ({probabilityRandom.DicOfProbs[item.Key]})");
sum += item.Value / maxRand;
}
console.WriteLine($"P = {sum}");
}
public (int, bool) Act()
{
if (Ran.NextDouble() < Epsilon)
{
return(Indices[Ran.Next(KArm)], false);
}
if (UcbParam.HasValue)
{
UcbEstimation = QEstimation.Select((data, index) =>
{
return(data + UcbParam.Value * Math.Sqrt(Math.Log(Time + 1) / (ActionCount[index] + (10 ^ -5))));
}).ToList();
QBest = UcbEstimation.Max();
var ucbEstMaxList = UcbEstimation.Where(v => v == QBest).Select((data, index) => new {
Index = index,
Data = data
}).ToList();
return(ucbEstMaxList[Ran.Next(ucbEstMaxList.Count())].Index, true);
}
if (Gradient)
{
ExpEst = QEstimation.Select(qEst => Math.Exp(qEst)).ToList();
ActionProb = ExpEst.Select(expEst => expEst / ExpEst.Sum()).ToList();
ProbabilityRandom.Reset();
for (int i = 0; i < ActionProb.Count; i++)
{
ProbabilityRandom.SetProb(i, ActionProb[i]);
}
//var acts = ActionProb.TakeWhile(p => Ran.NextDouble() < p).Select((d, i) => i).ToList();
//if (acts.Count == 0)
// return (ActionProb.Select((d, i) => i).ToList()[Ran.Next(ActionProb.Count)], false);
//return (acts[Ran.Next(acts.Count)], false);
return(ProbabilityRandom.Next(), false);
}
return(QEstimation.IndexOf(QEstimation.Max()), true);
}
