public static double[] denominateCoefficient(TSAFuzzySystem System, Matrix Teta)
{
double[] result = new double[Teta.RowDimension - 1];
for (int i = 1; i < Teta.RowDimension; i++)
{
result[i - 1] = (Teta.GetElement(i, 0)); // *System.LearnSamplesSet.Output_Attributes.Scatter) / System.LearnSamplesSet.Input_Attributes[i - 1].Scatter;
}
return(result);
}
public TSAFuzzySystem(TSAFuzzySystem Source)
: base(Source)
{
RulesDatabaseSet = new List <KnowlegeBaseTSARules>();
for (int i = 0; i < Source.RulesDatabaseSet.Count; i++)
{
RulesDatabaseSet.Add(new KnowlegeBaseTSARules(Source.RulesDatabaseSet[i]));
}
}
public static double[] denominateCoefficient(TSAFuzzySystem System, Matrix Teta)
{
double[] result = new double[Teta.RowDimension - 1];
for (int i = 1; i < Teta.RowDimension; i++)
{
result[i - 1] = (Teta.GetElement(i, 0));
}
return(result);
}
public override string ToString(bool with_param = false)
{
if (with_param)
{
string result = "Swallow Swarm Optimization{";
result += "}";
return(result);
}
return("Swallow Swarm Optimization");
}
public static double denominateIndependCoeff(TSAFuzzySystem System, Matrix Teta)
{
double result = Teta.GetElement(0, 0);
/* double sum = 0;
* for (int i = 1; i < Teta.RowDimension; i++)
* {
* sum = sum + (Teta.GetElement(i, 0) * System.LearnSamplesSet.InputAttributeMin(i - 1)) / System.LearnSamplesSet.InputAttributeScatter(i - 1);
* }
* result = result - sum;
* result *= System.LearnSamplesSet.Output_Attributes.Scatter;
* result += System.LearnSamplesSet.Output_Attributes.Min; */
return(result);
}
public override string ToString(bool with_param = false)
{
if (with_param)
{
string result = "Water cycle algorithm" + " {" + Environment.NewLine;
result += "Итераций= " + MaxIter.ToString() + " ;" + Environment.NewLine;
result += "Капель= " + Npop.ToString() + " ;" + Environment.NewLine;
result += "Рек= " + NRivers.ToString() + " ;" + Environment.NewLine;
result += "Dmax= " + Dmax.ToString() + " ;" + Environment.NewLine;
result += "const= " + con.ToString() + ";" + Environment.NewLine;
result += "Испарение - " + flag + "; }" + Environment.NewLine;
return(result);
}
return("Water cycle algorithm");
}
public static double EvaluteConsiquent(TSAFuzzySystem System, List <Term> termSet, out double[] RegressionCoefficents)
{
Matrix nominateVector, PnominatorMul1, PnominatorMul2, PnominatorMul3, UnidenominatorPart1, Unidenominator2, TetaResultPart1, Tetanominator1, TetaResultBr2Part1;
double TetaResultBr2, UniDenominator;
RegressionCoefficents = null;
int countParams = System.CountFeatures;// termSet.Count();
double a = Math.Pow(10, 100);
Matrix P = new Matrix(countParams + 1, countParams + 1, 0);
for (int i = 0; i < countParams + 1; i++)
{
P.SetElement(i, i, a);
}
Matrix Teta = new Matrix(countParams + 1, 1, 0);
List <SampleSet.RowSample> RealValue = SelectPoint(System, termSet, countParams);
List <SampleSet.RowSample> NominateValue = nominateValue(System, RealValue);
for (int i = 0; i < NominateValue.Count; i++)
{
nominateVector = new Matrix(NominateValue[i].InputAttributeValue, countParams + 1);
UnidenominatorPart1 = nominateVector.Transpose().Multiply(P);
Unidenominator2 = UnidenominatorPart1.Multiply(nominateVector);
UniDenominator = 1.0 / (1.0 / (calcWeigth(NominateValue[i], termSet)) + Unidenominator2.GetElement(0, 0));
Tetanominator1 = P.Multiply(nominateVector);
TetaResultPart1 = Tetanominator1.Multiply(UniDenominator);
TetaResultBr2Part1 = nominateVector.Transpose().Multiply(Teta);
TetaResultBr2 = NominateValue[i].DoubleOutput - TetaResultBr2Part1.GetElement(0, 0);
Teta = Teta + TetaResultPart1.Multiply(TetaResultBr2);
// Calc P
PnominatorMul1 = P.Multiply(nominateVector);
PnominatorMul2 = PnominatorMul1.Multiply(nominateVector.Transpose());
PnominatorMul3 = PnominatorMul2.Multiply(P);
P = P - PnominatorMul3.Multiply(UniDenominator);
//End
}
RegressionCoefficents = denominateCoefficient(System, Teta);
return(denominateIndependCoeff(System, Teta));
}
public void ConstructNewRule(List <Term> terms, TSAFuzzySystem FS)
{
TSARule Result;
int[] order_of_terms = new int[terms.Count()];
for (int i = 0; i < terms.Count(); i++)
{
order_of_terms[i] = TermsSet.Count;
TermsSet.Add(terms[i]);
}
double [] temp_regressionCoefficent;
double kons_Value = LSMWeghtReqursiveSimple.EvaluteConsiquent(FS, TermsSet, out temp_regressionCoefficent);
Result = new TSARule(TermsSet, order_of_terms, kons_Value, temp_regressionCoefficent);
RulesDatabase.Add(Result);
}
private static List <SampleSet.RowSample> MakeXValue(TSAFuzzySystem System, List <SampleSet.RowSample> Source)
{
List <SampleSet.RowSample> result = new List <SampleSet.RowSample>();
for (int i = 0; i < Source.Count; i++)
{
double[] inputValue = new double[Source[i].InputAttributeValue.Length + 1];
inputValue[0] = 1;
for (int j = 1; j < Source[i].InputAttributeValue.Length + 1; j++)
{
inputValue[j] = (Source[i].InputAttributeValue[j - 1]);
}
double Value = (Source[i].DoubleOutput);
result.Add(new SampleSet.RowSample(inputValue, null, Value, ""));
}
return(result);
}
public override TSAFuzzySystem TuneUpFuzzySystem(TSAFuzzySystem Approx, ILearnAlgorithmConf conf)
{
result = Approx;
Init(conf);
HeadLeader = new KnowlegeBaseTSARules(result.RulesDatabaseSet[0]);
VelocityVector = new KnowlegeBaseTSARules(result.RulesDatabaseSet[0]);
VelocityVectorLL = new KnowlegeBaseTSARules(result.RulesDatabaseSet[0]);
VelocityVectorHL = new KnowlegeBaseTSARules(result.RulesDatabaseSet[0]);
for (int i = 0; i < VelocityVector.TermsSet.Count; i++)
{
for (int j = 0; j < VelocityVector.TermsSet[i].Parametrs.Length; j++)
{
VelocityVector.TermsSet[i].Parametrs[j] = 0;
VelocityVectorLL.TermsSet[i].Parametrs[j] = 0;
VelocityVectorHL.TermsSet[i].Parametrs[j] = 0;
}
}
SetPopulation();
ParticlesBest = new Dictionary <KnowlegeBaseTSARules, KnowlegeBaseTSARules>();
foreach (var Particle in Population)
{
ParticlesBest.Add(Particle, Universal);
}
LocalLeaders = new KnowlegeBaseTSARules[numberOfLocalLeaders];
ExplorerParticles = new KnowlegeBaseTSARules[numberOfAllParts - numberOfAimlessParts - numberOfLocalLeaders - 1];
AimlessParticles = new KnowlegeBaseTSARules[numberOfAimlessParts];
iter = 0;
while (iter < MaxIter)
{
Population = ListTakagiSugenoApproximateTool.SortRules(Population, result);
SetRoles();
ChangeExplorersPositions();
ChangeAimlessPositions();
DiscardRoles();
if (iter == (MaxIter - 1) || iter == 0)
{
Console.WriteLine("Iteration: " + (iter + 1).ToString());
Console.WriteLine(Math.Round(result.RMSEtoMSEforLearn(result.approxLearnSamples(Population[0])), 3));
Console.WriteLine(Math.Round(result.RMSEtoMSEforTest(result.approxTestSamples(Population[0])), 3));
}
iter++;
}
result.RulesDatabaseSet[0] = Population[0];
return(result);
}
private static List <SampleSet.RowSample> nominateValue(TSAFuzzySystem System, List <SampleSet.RowSample> Source)
{
List <SampleSet.RowSample> result = new List <SampleSet.RowSample>();
for (int i = 0; i < Source.Count; i++)
{
double[] inputValue = new double[Source[i].InputAttributeValue.Length + 1];
inputValue[0] = 1;
for (int j = 1; j < Source[i].InputAttributeValue.Length + 1; j++)
{
inputValue[j] = (Source[i].InputAttributeValue[j - 1]); // - System.LearnSamplesSet.InputAttributeMin(j - 1) ) / (System.LearnSamplesSet.InputAttributeScatter(j - 1));
}
double Value = (Source[i].DoubleOutput); // - System.LearnSamplesSet.Output_Attributes.Min) / System.LearnSamplesSet.Output_Attributes.Scatter;
result.Add(new SampleSet.RowSample(inputValue, null, Value, ""));
}
return(result);
}
public static List <SampleSet.RowSample> SelectPoint(TSAFuzzySystem System, List <Term> termSet, int CountParams, double SelectLevel = 0.000000001)
{
List <SampleSet.RowSample> result = new List <SampleSet.RowSample>();
List <double> Distance = new List <double>();
double[] min = new double[termSet.Count];
double[] max = new double[termSet.Count];
double[] temp;
for (int i = 0; i < termSet.Count; i++)
{
temp = termSet[i].getXValueByLevelMembership(SelectLevel);
min[termSet[i].NumVar] = temp[0];
max[termSet[i].NumVar] = temp[1];
}
for (int i = 0; i < System.LearnSamplesSet.CountSamples; i++)
{
bool flag = true;
for (int j = 0; j < termSet.Count; j++)
{
if (System.AcceptedFeatures[termSet[j].NumVar] == false)
{
continue;
}
if ((System.LearnSamplesSet.DataRows[i].InputAttributeValue[termSet[j].NumVar] <= min[termSet[j].NumVar]) &&
(System.LearnSamplesSet.DataRows[i].InputAttributeValue[termSet[j].NumVar] >= max[termSet[j].NumVar]))
{
flag = false;
break;
}
}
if (flag)
{
result.Add(System.LearnSamplesSet[i]);
}
}
return(result);
}
public override TSAFuzzySystem TuneUpFuzzySystem(TSAFuzzySystem Approximate, ILearnAlgorithmConf Conf)
{
result = Approximate;
List <int[]> groups = new List <int[]>();
Init(Conf);
SetPopulation();
Population = SortRules(Population);
NS = new int[Nsr];
NS = SetNS(Population, Nsr);
groups = GroupStream();
double BestMSETest = result.RMSEtoMSEforTest(result.approxTestSamples(Population[0]));
double BestMSELearn = result.RMSEtoMSEforLearn(result.approxLearnSamples(Population[0]));
int BestIter = 0;
for (int i = 1; i <= MaxIter; i++)
{
Console.WriteLine(i + " - Итерация");
Population = SetNextPosition(groups, Population);
Population = Replacement(groups, Population);
if (flag)
{
Evaporation(groups.Last());//Испарение
}
if (BestMSETest > result.RMSEtoMSEforTest(result.approxTestSamples(Population[0])))
{
BestMSETest = result.RMSEtoMSEforTest(result.approxTestSamples(Population[0]));
BestMSELearn = result.RMSEtoMSEforLearn(result.approxLearnSamples(Population[0]));
BestIter = i;
}
}
Console.WriteLine(ToString(true));
Console.WriteLine("Итер - " + BestIter + " MSET - " + BestMSETest + " MSEL - " + BestMSELearn);
result.RulesDatabaseSet[0] = Population[0];
return(result);
}
public override FuzzySystem.TakagiSugenoApproximate.TSAFuzzySystem Generate(FuzzySystem.TakagiSugenoApproximate.TSAFuzzySystem Approximate, FuzzySystem.FuzzyAbstract.conf.IGeneratorConf config)
{
TSAFuzzySystem result = Approximate;
type_func = TypeTermFuncEnum.Гауссоида;
ResultSystem = Approximate;
var kliConf = config as KLI_conf;
if (kliConf != null)
{
double meanValue = result.LearnSamplesSet.DataRows.Select(x => x.DoubleOutput).Average();
var mayError = kliConf.MaxValue * meanValue;
double centerValue = mayError;
ScoreListElem[] ResultsKLI = new ScoreListElem[3];
ScoreListElem[] ResultsKLI2 = new ScoreListElem[3];
List <ScoreListElem> ResultsKLIFull = new List <ScoreListElem>();
List <ScoreListElem> ResultsKLI2Full = new List <ScoreListElem>();
double magic = 0.25;
KLI.kliGenerate(result, type_func, centerValue);
ScoreListElem tempRes = new ScoreListElem();
tempRes.MaxError = centerValue;
tempRes.Alg = "Кли";
tempRes.Res = result.RulesDatabaseSet[0];
tempRes.Error = result.ErrorLearnSamples(tempRes.Res);
ResultsKLI[1] = new ScoreListElem(tempRes);
if (!double.IsNaN(tempRes.Error))
{
ResultsKLIFull.Add(new ScoreListElem(tempRes));
}
for (int i = 0; i < 100; i++)
{
double leftValue = centerValue * (1 - magic);
double rigthValue = centerValue * (1 + magic);
tempRes = new ScoreListElem();
KLI.kliGenerate(result, type_func, leftValue);
tempRes.Alg = "Кли";
tempRes.MaxError = leftValue;
tempRes.Res = result.RulesDatabaseSet[0];
tempRes.Error = result.ErrorLearnSamples(tempRes.Res);
ResultsKLI[0] = new ScoreListElem(tempRes);
if (!double.IsNaN(tempRes.Error))
{
ResultsKLIFull.Add(new ScoreListElem(tempRes));
}
result.RulesDatabaseSet.Clear();
tempRes = new ScoreListElem();
KLI.kliGenerate(result, type_func, rigthValue);
tempRes.Alg = "Кли";
tempRes.MaxError = rigthValue;
tempRes.Res = result.RulesDatabaseSet[0];
tempRes.Error = result.ErrorLearnSamples(tempRes.Res);
ResultsKLI[2] = tempRes;
if (!double.IsNaN(tempRes.Error))
{
ResultsKLIFull.Add(new ScoreListElem(tempRes));
}
result.RulesDatabaseSet.Clear();
if (((ResultsKLI[0].Error >= ResultsKLI[1].Error) || double.IsNaN(ResultsKLI[0].Error)) && ((ResultsKLI[1].Error <= ResultsKLI[2].Error) || (double.IsNaN(ResultsKLI[2].Error))))
{
magic *= 0.7;
}
double err = ResultsKLIFull.Min(y => y.Error);
ResultsKLI[1].Error = err;
centerValue = ResultsKLIFull.First(x => x.Error == err).MaxError;
}
centerValue = mayError;
tempRes = new ScoreListElem();
kliGenerate(result, type_func, centerValue);
tempRes.Alg = "Кли2";
tempRes.MaxError = centerValue;
tempRes.Res = result.RulesDatabaseSet[0];
tempRes.Error = result.ErrorLearnSamples(tempRes.Res);
ResultsKLI2[1] = new ScoreListElem(tempRes);
if (!double.IsNaN(tempRes.Error))
{
ResultsKLI2Full.Add(new ScoreListElem(tempRes));
}
result.RulesDatabaseSet.Clear();
magic = 0.25;
for (int i = 0; i < 100; i++)
{
double leftValue = centerValue * (1 - magic);
double rigthValue = centerValue * (1 + magic);
tempRes = new ScoreListElem();
kliGenerate(result, type_func, leftValue);
tempRes.Alg = "Кли2";
tempRes.MaxError = leftValue;
tempRes.Res = result.RulesDatabaseSet[0];
tempRes.Error = result.ErrorLearnSamples(tempRes.Res);
ResultsKLI2[0] = new ScoreListElem(tempRes);
if (!double.IsNaN(tempRes.Error))
{
ResultsKLI2Full.Add(new ScoreListElem(tempRes));
}
result.RulesDatabaseSet.Clear();
tempRes = new ScoreListElem();
kliGenerate(result, type_func, rigthValue);
tempRes.Alg = "Кли2";
tempRes.MaxError = rigthValue;
tempRes.Res = result.RulesDatabaseSet[0];
tempRes.Error = result.ErrorLearnSamples(tempRes.Res);
ResultsKLI2[2] = new ScoreListElem(tempRes);
if (!double.IsNaN(tempRes.Error))
{
ResultsKLI2Full.Add(new ScoreListElem(tempRes));
}
result.RulesDatabaseSet.Clear();
if (((ResultsKLI2[0].Error >= ResultsKLI2[1].Error) || double.IsNaN(ResultsKLI2[0].Error)) && ((ResultsKLI2[1].Error <= ResultsKLI2[2].Error) || (double.IsNaN(ResultsKLI2[2].Error))))
{
magic *= 0.7;
}
double err = ResultsKLI2Full.Min(y => y.Error);
ResultsKLI2[1].Error = err;
centerValue = ResultsKLI2Full.First(x => x.Error == err).MaxError;
}
ScoreListElem BestKli = ResultsKLIFull.Where(x => x.Error == ResultsKLIFull.Min(y => y.Error)).First();
ScoreListElem BestKli2 = ResultsKLI2Full.Where(x => x.Error == ResultsKLI2Full.Min(y => y.Error)).First();
if (BestKli.Error > BestKli2.Error)
{
Best = BestKli2;
}
else
{
Best = BestKli;
}
result.RulesDatabaseSet.Add(Best.Res);
}
return(result);
}
public static List <SampleSet.RowSample> SelectPoint(TSAFuzzySystem System, List <Term> termSet, int CountParams, double SelectLevel = 0.000000001)
{
List <SampleSet.RowSample> result = new List <SampleSet.RowSample>();
List <double> Distance = new List <double>();
double[] min = new double[termSet.Count];
double[] max = new double[termSet.Count];
double[] temp;
for (int i = 0; i < termSet.Count; i++)
{
temp = termSet[i].getXValueByLevelMembership(SelectLevel);
min[termSet[i].NumVar] = temp[0];
max[termSet[i].NumVar] = temp[1];
}
for (int i = 0; i < System.LearnSamplesSet.CountSamples; i++)
{
bool flag = true;
for (int j = 0; j < System.LearnSamplesSet.CountVars; j++)
{
if (System.AcceptedFeatures[j] == false)
{
continue;
}
if ((System.LearnSamplesSet.DataRows[i].InputAttributeValue[j] <= min[j]) && (System.LearnSamplesSet.DataRows[i].InputAttributeValue[j] >= max[j]))
{
flag = false;
break;
}
}
if (flag)
{
result.Add(System.LearnSamplesSet[i]);
}
}
if (result.Count <= CountParams)
{
for (int c = 0; c < System.LearnSamplesSet.CountSamples; c++)
{
double current_diff = 0;
for (int i = 0; i < termSet.Count; i++)
{
if (System.AcceptedFeatures[termSet[i].NumVar] == false)
{
continue;
}
if (termSet[i] != null)
{
switch (termSet[i].TermFuncType)
{
case TypeTermFuncEnum.Треугольник:
current_diff +=
Math.Abs(System.LearnSamplesSet.DataRows[c].InputAttributeValue[termSet[i].NumVar] -
termSet[i].Parametrs[1]);
break;
case TypeTermFuncEnum.Гауссоида: current_diff +=
Math.Abs(System.LearnSamplesSet.DataRows[c].InputAttributeValue[termSet[i].NumVar] -
termSet[i].Parametrs[0]);
break;
case TypeTermFuncEnum.Парабола:
double argv = (termSet[i].Parametrs[0] + termSet[i].Parametrs[1]) / 2;
current_diff +=
Math.Abs(System.LearnSamplesSet.DataRows[c].InputAttributeValue[termSet[i].NumVar] -
argv);
break;
case TypeTermFuncEnum.Трапеция: double argvTR = (termSet[i].Parametrs[1] + termSet[i].Parametrs[2]) / 2;
current_diff +=
Math.Abs(System.LearnSamplesSet.DataRows[c].InputAttributeValue[termSet[i].NumVar] -
argvTR);
break;
}
}
}
if (result.Contains(System.LearnSamplesSet.DataRows[c]))
{
Distance.Add(double.PositiveInfinity);
}
else
{
Distance.Add(current_diff);
}
}
while (result.Count <= CountParams)
{
double cur_min = Distance.Min();
int iMinIndex = Distance.IndexOf(cur_min);
result.Add(System.LearnSamplesSet.DataRows[iMinIndex]);
Distance[iMinIndex] = double.PositiveInfinity;
}
}
return(result);
}
public static double EvaluteConsiquent(TSAFuzzySystem System, TermSetGlobal <Term> termSet, out double[] RegressionCoefficents)
{
return(EvaluteConsiquent(System, termSet.ToList(), out RegressionCoefficents));
}
