source: trunk/yat/utility/NNI.cc @ 1615

Last change on this file since 1615 was 1487, checked in by Jari Häkkinen, 13 years ago

Addresses #436. GPL license copy reference should also be updated.

  • Property svn:eol-style set to native
  • Property svn:keywords set to Author Date Id Revision
File size: 3.1 KB
Line 
1// $Id: NNI.cc 1487 2008-09-10 08:41:36Z jari $
2
3/*
4  Copyright (C) 2004 Jari Häkkinen
5  Copyright (C) 2005 Peter Johansson
6  Copyright (C) 2006 Jari Häkkinen
7  Copyright (C) 2007 Jari Häkkinen, Peter Johansson
8  Copyright (C) 2008 Peter Johansson
9
10  This file is part of the yat library, http://dev.thep.lu.se/yat
11
12  The yat library is free software; you can redistribute it and/or
13  modify it under the terms of the GNU General Public License as
14  published by the Free Software Foundation; either version 3 of the
15  License, or (at your option) any later version.
16
17  The yat library is distributed in the hope that it will be useful,
18  but WITHOUT ANY WARRANTY; without even the implied warranty of
19  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20  General Public License for more details.
21
22  You should have received a copy of the GNU General Public License
23  along with yat. If not, see <http://www.gnu.org/licenses/>.
24*/
25
26#include "NNI.h"
27#include "stl_utility.h"
28
29#include <algorithm>
30#include <cmath>
31#include <fstream>
32
33namespace theplu {
34namespace yat {
35namespace utility {
36
37  // For a discussion and motivation for various algorithm
38  // implementations here see the paper cited in the class definition
39  // documentation.
40  NNI::NNI(const utility::Matrix& matrix,const utility::Matrix& weight,
41           const unsigned int neighbours)
42    : data_(matrix), imputed_data_(matrix), neighbours_(neighbours),
43      weight_(weight)
44  {
45  }
46
47
48  // d_{ij}^2=\frac {\sum_{k=1,C} w_{ik} w_{jk} (x_{ik}-x_{jk})^2 }
49  //                {\sum_{k=l,C} w_{ik} w_{jk} }
50  // where C is the number of columns
51  std::vector<std::pair<size_t,double> >
52  NNI::calculate_distances(const size_t row) const
53  {
54    std::vector<std::pair<size_t,double> > distance;
55    for (size_t i=0; i<data_.rows(); i++)
56      if (i!=row) {
57        double contribs=0;
58        std::pair<size_t,double> this_distance(i,0.0);
59        for (size_t j=0; j<data_.columns(); j++)
60          // 0 contribution for missing values
61          if (weight_(i,j) && weight_(row,j)) {
62            double weight_factor=weight_(i,j)*weight_(row,j);
63            this_distance.second+=( weight_factor *
64                                    (data_(i,j)-data_(row,j)) *
65                                    (data_(i,j)-data_(row,j)) );
66            contribs+=weight_factor;
67          }
68        if (contribs) { // ignore lines without any contributions
69          this_distance.second=sqrt(this_distance.second/contribs);
70          distance.push_back(this_distance);
71        }
72      }
73    return distance;
74  }
75
76
77  const utility::Matrix& NNI::imputed_data(void) const
78  {
79    return imputed_data_;
80  }
81
82
83  const std::vector<size_t>& NNI::not_imputed(void) const
84  {
85    return not_imputed_;
86  }
87
88
89  // Contributing nearest neighbours are added up to the user set
90  // number, and neighbours are disqualified if their element (column)
91  // weight is zero
92  std::vector<size_t> 
93  NNI::nearest_neighbours(const size_t column,
94                          const std::vector<std::pair<size_t,double> >& d) const
95  {
96    std::vector<size_t> index;
97    double contribs=0;
98    for (size_t i=0; ((i<d.size()) &&
99                     (contribs+=weight_(d[i].first,column))<=neighbours_); i++)
100      if (weight_(d[i].first,column))
101        index.push_back(i);
102    return index;
103  }
104
105}}} // of namespace utility, yat, and theplu
Note: See TracBrowser for help on using the repository browser.