source: trunk/yat/regression/Local.cc @ 1486

// $Id: Local.cc 1486 2008-09-09 21:17:19Z jari $

/*
  Copyright (C) 2004 Peter Johansson
  Copyright (C) 2005, 2006, 2007 Jari Häkkinen, Peter Johansson
  Copyright (C) 2008 Peter Johansson

  This file is part of the yat library, http://dev.thep.lu.se/yat

  The yat library is free software; you can redistribute it and/or
  modify it under the terms of the GNU General Public License as
  published by the Free Software Foundation; either version 3 of the
  License, or (at your option) any later version.

  The yat library is distributed in the hope that it will be useful,
  but WITHOUT ANY WARRANTY; without even the implied warranty of
  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  General Public License for more details.

  You should have received a copy of the GNU General Public License
  along with this program; if not, write to the Free Software
  Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
  02111-1307, USA.
*/

#include "Local.h"
#include "Kernel.h"
#include "OneDimensionalWeighted.h"
#include "yat/utility/Vector.h"
#include "yat/utility/VectorView.h"

#include <algorithm>
#include <cassert>
#include <iostream>
#include <sstream>
#include <stdexcept>

namespace theplu {
namespace yat {
namespace regression {

  Local::Local(OneDimensionalWeighted& r, Kernel& k)
    : kernel_(&k), regressor_(&r)
  {
  }

  Local::~Local(void)
  {
  }

  void Local::add(const double x, const double y)
  {
    data_.push_back(std::make_pair(x,y));
  }

  void Local::fit(const size_t step_size, const size_t nof_points)
  {
    if (step_size==0){
      std::stringstream ss;
      ss << "yat::regression::Local: step_size must be larger than zero.";
      throw std::runtime_error(ss.str());
    }
    if (nof_points<3){
      std::stringstream ss;
      ss << "yat::regression::Local: too few data points. "
         << "At least 3 data points are needed to perform fitting.";
      throw std::runtime_error(ss.str());
    }
    if (data_.size()<step_size){
      std::stringstream ss;
      ss << "yat::regression::Local: too large step_size "
         << "step_size, " << step_size 
         << ", is larger than number of added data points " << data_.size();
      throw std::runtime_error(ss.str());
    }

    size_t nof_fits=data_.size()/step_size;
    x_.resize(nof_fits);
    y_predicted_.resize(x_.size());
    y_err_.resize(x_.size());
    sort(data_.begin(), data_.end());

    // copying data to 2 utility vectors ONCE to use views from
    utility::Vector x(data_.size());
    utility::Vector y(data_.size());
    for (size_t j=0; j<x.size(); j++){
      x(j)=data_[j].first;
      y(j)=data_[j].second;
    }

    // looping over regression points and perform local regression
    for (size_t i=0; i<nof_fits; i++) {
      size_t max_index = static_cast<size_t>( (i+0.5)*step_size );
      size_t min_index;
      double width; // distance from middle of windo to border of window
      double x_mid; // middle of window
      // right border case
      if (max_index > data_.size()-1){
        min_index = max_index - nof_points + 1;
        max_index = data_.size()-1;
        width = ( (( x(max_index)-x(0) )*(nof_points-1)) / 
                  ( 2*(max_index-min_index)) );
        x_mid = x(min_index)+width;
      }
      // normal middle case
      else if (max_index > nof_points-1){
        min_index = max_index - nof_points + 1;
        width = (x(max_index)-x(min_index))/2;
        x_mid = x(min_index)+width;
      }
      // left border case
      else {
        min_index = 0;
        width = ( (( x(max_index)-x(0) )*(nof_points-1)) / 
                  ( 2*(max_index-min_index)) );
        x_mid = x(max_index)-width;
      }
      assert(min_index<data_.size());
      assert(max_index<data_.size());
                                
      utility::VectorView x_local(x, min_index, max_index-min_index+1);
      utility::VectorView y_local(y, min_index, max_index-min_index+1);

      // calculating weights
      utility::Vector w(max_index-min_index+1);
      for (size_t j=0; j<w.size(); j++)
        w(j) = (*kernel_)( (x_local(j)- x_mid)/width );
      
      // fitting the regressor locally
      regressor_->fit(x_local,y_local,w);
      assert(i<y_predicted_.size());
      assert(i<y_err_.size());
      x_(i) = x(i*step_size);
      y_predicted_(i) = regressor_->predict(x(i*step_size));
      y_err_(i) = sqrt(regressor_->standard_error2(x(i*step_size)));
    }
  }


  void Local::reset(void)
  {
    data_.clear();
    x_.resize(0);
    y_predicted_.resize(0);
    y_err_.resize(0);
  }


  const utility::Vector& Local::x(void) const
  {
    return x_;
  }

  const utility::Vector& Local::y_predicted(void) const
  {
    return y_predicted_;
  }

  const utility::Vector& Local::y_err(void) const
  {
    return y_err_;
  }

  std::ostream& operator<<(std::ostream& os, const Local& r)
  {
    os << "# column 1: x\n"
      << "# column 2: y\n"
      << "# column 3: y_err\n";
    for (size_t i=0; i<r.x().size(); i++) {
      os << r.x()(i) << "\t" 
         << r.y_predicted()(i) << "\t"
         << r.y_err()(i) << "\n";
    }   

    return os;
  }

}}} // of namespaces regression, yat, and theplu