rsvm_surr_strategy.hpp

 
#include <libcmaes/cmaes.h>
#include <libcmaes/surrogatestrategy.h>
#include <libcmaes/opti_err.h>
#include <libcmaes/surrogates/rankingsvm.hpp>
 
#ifndef RSVMSURROGATESTRATEGY_H
#define RSVMSURROGATESTRATEGY_H
 
namespace libcmaes
{
 
  // Interfacing candidates with an Eigen matrix of double x and a vector of objective function value.
  void to_mat_vec(std::vector<Candidate> &cp,
          dMat &x, dVec &fvalues,
          const bool &train)
  {
    if (train)
      std::sort(cp.begin(),cp.end(),
        [](Candidate const &c1, Candidate const &c2){return c1.get_fvalue() > c2.get_fvalue();}); // descending sort
    x = dMat(cp.at(0).get_x_size(),cp.size());
    fvalues = dVec(cp.size());
    for (int i=0;i<(int)cp.size();i++)
      {
    x.col(i) = cp.at(i).get_x_dvec().transpose();
    fvalues(i) = cp.at(i).get_fvalue();
      }
  }
 
  template <class TGenoPheno> using eostrat = ESOStrategy<CMAParameters<TGenoPheno>,CMASolutions,CMAStopCriteria<TGenoPheno> >;
 
  template<template <class U, class V> class TStrategy, class TCovarianceUpdate=CovarianceUpdate,class TGenoPheno=GenoPheno<NoBoundStrategy>>
  class RSVMSurrogateStrategy : public ACMSurrogateStrategy<TStrategy,TCovarianceUpdate,TGenoPheno>
    {
    public:
    RSVMSurrogateStrategy(FitFunc &func,
              CMAParameters<TGenoPheno> &parameters)
          :ACMSurrogateStrategy<TStrategy,TCovarianceUpdate,TGenoPheno>(func,parameters)
    {
      this->_train = [this](const std::vector<Candidate> &c, const dMat &cov)
      {
    if (c.empty())
      return 0;
    dMat x;
    dVec fvalues;
    std::vector<Candidate> cp = c;
    to_mat_vec(cp,x,fvalues,true);
    dVec xmean = eostrat<TGenoPheno>::get_solutions().xmean();
    _rsvm = RankingSVM<RBFKernel>();
    _rsvm._encode = true;
    _rsvm.train(x,_rsvm_iter,cov,xmean);
    return 0;
      };
      this->_predict = [this](std::vector<Candidate> &c, const dMat &cov)
      {
    dMat x_test(c.at(0).get_x_size(),c.size());
    for (int i=0;i<(int)c.size();i++)
      x_test.col(i) = c.at(i).get_x_dvec().transpose();
 
    dMat x_train;
    dVec fvalues;
    std::vector<Candidate> tset = this->_tset;
    to_mat_vec(tset,x_train,fvalues,true);
 
    dVec fit;
    dVec xmean = eostrat<TGenoPheno>::get_solutions().xmean();
    _rsvm.predict(fit,x_test,x_train,cov,xmean);
    if (fit.size() != 0)
      for (int i=0;i<(int)c.size();i++)
        c.at(i).set_fvalue(fit(i));
    return 0;
      };
    }
 
      ~RSVMSurrogateStrategy() {}
 
      RankingSVM<RBFKernel> _rsvm;
      int _rsvm_iter = 1e6; 
  };
 
}
 
#endif