pso.cpp

#include "pso.h"

PSO::PSO()
{
    this->varNum = 1;
    this->PopSize = 80;
    this->GenSize = 2000;
    this->phi1 = 1;
    this->phi2 = 1;
    this->omega = 1;
    this->MinMax = false;
}

PSO::PSO(int varNum, int PopSize, int GenSize)
{

    this->varNum = varNum;
    this->PopSize = PopSize;
    this->GenSize = GenSize;
    this->phi1 = 1;
    this->phi2 = 1;
    this->omega = 1;
    this->MinMax = false;

}

PSO::PSO(int varNum, int PopSize, int GenSize, double phi1, double phi2)
{
    this->varNum = varNum;
    this->PopSize = PopSize;
    this->GenSize = GenSize;
    this->phi1 = phi1;
    this->phi2 = phi2;
    this->omega = 1;
    this->MinMax = false;
}

PSO::PSO(int varNum, int PopSize, int GenSize, double phi1, double phi2, double omega, bool MinMax)
{
    this->varNum = varNum;
    this->PopSize = PopSize;
    this->GenSize = GenSize;
    this->phi1 = phi1;
    this->phi2 = phi2;
    this->omega = omega;
    this->MinMax = MinMax;
}

PSO::~PSO()
{

}

void PSO::initAlgorithm()
{
    this->X.randU(this->PopSize, this->varNum);
    this->X = this->X*(4.0)-2.0;
    this->V.randU(this->PopSize, this->varNum);
    this->P = this->X;
    this->G.randU(1, this->varNum);
    this->Xfitness = Evaluation(this->X);
//    Xfitness.print();
    this->Gfitness = Evaluation(this->G);
//    Gfitness.print();
    this->Pfitness = this->Xfitness;

}

Matrix PSO::Evaluation(Matrix Matrix2Evaluate)
{
    Matrix ret(Matrix2Evaluate.getRows(), 1), EstimatedOutPut;
    Lsim ArxModel;

    ArxModel.addIO(this->dataIn, this->dataOut);
    //TODO -> Tornar a função mais fléxivel.
    for(int i = 1; i <= Matrix2Evaluate.getRows(); i++)
    {
      EstimatedOutPut = ArxModel.simArxOneStep(2, 2, Matrix2Evaluate.getLine(i));
      ret.add(i, 1, norm(this->dataOut - EstimatedOutPut));
    }


    return ret;

}

void PSO::ParticleUpdate()
{
    this->X = this->X + this->V;
}

void PSO::VelocityUpdate()
{
    double Rand1, Rand2;
    srand(time(NULL));

    for(int i = 1; i <= this->PopSize; i++)
        for(int j = 1; j <= this->varNum; j++)
        {
            Rand1 = random()*this->phi1;
            Rand2 = random()*this->phi2;
            this->V.add(i, j, (this->omega*this->V(i,j) + Rand1*(this->P(i,j) - this->X(i,j)) + Rand2*(this->G(1,j) - this->X(i,j))));
        }
}

void PSO::ParticleEvaluation()
{
    this->Xfitness = Evaluation(this->X);
}

void PSO::FitnessUpdate()
{
    if(this->MinMax == true)
        FitnessUpdateMax();
    else
        FitnessUpdateMin();
}

void PSO::FitnessUpdateMin()
{
  for (int i = 1; i <= this->PopSize; i++)
  {
      if (this->Xfitness(i,1) < this->Pfitness(i,1))
      {
          this->Pfitness.add(i,1, this->Xfitness(i,1));
          for(int j = 1; j <= this->varNum; j++)
            this->P.add(i, j, this->X(i,j));
      }

      if (this->Pfitness(i,1) < this->Gfitness(1,1))
      {
          this->Gfitness.add(1,1, this->Pfitness(i,1));
          for(int j = 1; j <= this->varNum; j++)
            this->G.add(1, j, this->P(i,j));
      }
  }
}

void PSO::FitnessUpdateMax()
{
    for (int i = 1; i <= this->PopSize; i++)
    {
        if (this->Xfitness(i,1) > this->Pfitness(i,1))
        {
            this->Pfitness.add(i,1, this->Xfitness(i,1));
            for(int j = 1; j <= this->varNum; j++)
              this->P.add(i, j, this->X(i,j));
        }

        if (this->Pfitness(i,1) > this->Gfitness(1,1))
        {
            this->Gfitness.add(1,1, this->Pfitness(i,1));
            for(int j = 1; j <= this->varNum; j++)
              this->G.add(1, j, this->P(i,j));
        }
    }
}

void PSO::Run()
{
    initAlgorithm();

    for(int i = 0; i < this->GenSize; i++)
    {
        VelocityUpdate();
        ParticleUpdate();
        this->Xfitness = Evaluation(this->X);
        FitnessUpdate();
        this->GGen = this->GGen||this->G;
        this->GfitnessGen = this->GfitnessnTime||this->Gfitness;
    }    
}

void PSO::Run(int nTimes)
{
    for (int i = 0; i < nTimes; i++)
    {
        Run();
        this->GnTimes = this->GnTimes||this->G;
        this->GfitnessnTime =this->GfitnessnTime||this->Gfitness;
    }
}

void PSO::setData(Matrix dataIn, Matrix dataOut)
{
    this->dataIn = dataIn;
    this->dataOut = dataOut;
}