1 | #ifndef _theplu_yat_classifier_ensemblebuilder_ |
---|
2 | #define _theplu_yat_classifier_ensemblebuilder_ |
---|
3 | |
---|
4 | // $Id$ |
---|
5 | |
---|
6 | /* |
---|
7 | Copyright (C) 2005 Markus Ringnér |
---|
8 | Copyright (C) 2006 Jari Häkkinen, Markus Ringnér, Peter Johansson |
---|
9 | Copyright (C) 2007, 2008 Peter Johansson |
---|
10 | |
---|
11 | This file is part of the yat library, http://trac.thep.lu.se/yat |
---|
12 | |
---|
13 | The yat library is free software; you can redistribute it and/or |
---|
14 | modify it under the terms of the GNU General Public License as |
---|
15 | published by the Free Software Foundation; either version 2 of the |
---|
16 | License, or (at your option) any later version. |
---|
17 | |
---|
18 | The yat library is distributed in the hope that it will be useful, |
---|
19 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
---|
20 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
---|
21 | General Public License for more details. |
---|
22 | |
---|
23 | You should have received a copy of the GNU General Public License |
---|
24 | along with this program; if not, write to the Free Software |
---|
25 | Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA |
---|
26 | 02111-1307, USA. |
---|
27 | */ |
---|
28 | |
---|
29 | #include "FeatureSelector.h" |
---|
30 | #include "Sampler.h" |
---|
31 | #include "SubsetGenerator.h" |
---|
32 | #include "yat/statistics/Averager.h" |
---|
33 | #include "yat/utility/Matrix.h" |
---|
34 | |
---|
35 | #include <vector> |
---|
36 | |
---|
37 | namespace theplu { |
---|
38 | namespace yat { |
---|
39 | namespace classifier { |
---|
40 | |
---|
41 | /// |
---|
42 | /// @brief Class for ensembles of supervised classifiers |
---|
43 | /// |
---|
44 | template <class Classifier, class Data> |
---|
45 | class EnsembleBuilder |
---|
46 | { |
---|
47 | public: |
---|
48 | /** |
---|
49 | \brief Type of classifier that ensemble is built on. |
---|
50 | */ |
---|
51 | typedef Classifier classifier_type; |
---|
52 | |
---|
53 | /** |
---|
54 | Type of container used for storing data e.g. MatrixLookup or KernelLookup |
---|
55 | */ |
---|
56 | typedef Data data_type; |
---|
57 | |
---|
58 | /// |
---|
59 | /// Constructor. |
---|
60 | /// |
---|
61 | EnsembleBuilder(const Classifier&, const Data&, const Sampler&); |
---|
62 | |
---|
63 | /// |
---|
64 | /// Constructor. |
---|
65 | /// |
---|
66 | EnsembleBuilder(const Classifier&, const Data&, const Sampler&, |
---|
67 | FeatureSelector&); |
---|
68 | |
---|
69 | /// |
---|
70 | /// Destructor. |
---|
71 | /// |
---|
72 | virtual ~EnsembleBuilder(void); |
---|
73 | |
---|
74 | /// |
---|
75 | /// Generate ensemble. Function trains each member of the Ensemble. |
---|
76 | /// |
---|
77 | void build(void); |
---|
78 | |
---|
79 | /// |
---|
80 | /// @return classifier |
---|
81 | /// |
---|
82 | const Classifier& classifier(size_t i) const; |
---|
83 | |
---|
84 | /// |
---|
85 | /// @return Number of classifiers in ensemble |
---|
86 | /// |
---|
87 | u_long size(void) const; |
---|
88 | |
---|
89 | /// |
---|
90 | /// @brief Generate validation data for ensemble |
---|
91 | /// |
---|
92 | /// validate()[i][j] return averager for class @a i for sample @a j |
---|
93 | /// |
---|
94 | const std::vector<std::vector<statistics::Averager> >& validate(void); |
---|
95 | |
---|
96 | /** |
---|
97 | Predict a dataset using the ensemble. |
---|
98 | |
---|
99 | If @a data is a KernelLookup each column should correspond to a |
---|
100 | test sample and each row should correspond to a training |
---|
101 | sample. More exactly row \f$ i \f$ in @a data should correspond |
---|
102 | to the same sample as row/column \f$ i \f$ in the training |
---|
103 | kernel corresponds to. |
---|
104 | */ |
---|
105 | void predict(const Data& data, |
---|
106 | std::vector<std::vector<statistics::Averager> > &); |
---|
107 | |
---|
108 | private: |
---|
109 | // no copying |
---|
110 | EnsembleBuilder(const EnsembleBuilder&); |
---|
111 | const EnsembleBuilder& operator=(const EnsembleBuilder&); |
---|
112 | |
---|
113 | |
---|
114 | const Classifier& mother_; |
---|
115 | SubsetGenerator<Data>* subset_; |
---|
116 | std::vector<Classifier*> classifier_; |
---|
117 | std::vector<std::vector<statistics::Averager> > validation_result_; |
---|
118 | |
---|
119 | }; |
---|
120 | |
---|
121 | |
---|
122 | // implementation |
---|
123 | |
---|
124 | template <class C, class D> |
---|
125 | EnsembleBuilder<C,D>::EnsembleBuilder(const C& sc, const D& data, |
---|
126 | const Sampler& sampler) |
---|
127 | : mother_(sc),subset_(new SubsetGenerator<D>(sampler,data)) |
---|
128 | { |
---|
129 | } |
---|
130 | |
---|
131 | |
---|
132 | template <class C, class D> |
---|
133 | EnsembleBuilder<C, D>::EnsembleBuilder(const C& sc, const D& data, |
---|
134 | const Sampler& sampler, |
---|
135 | FeatureSelector& fs) |
---|
136 | : mother_(sc), |
---|
137 | subset_(new SubsetGenerator<D>(sampler,data,fs)) |
---|
138 | { |
---|
139 | } |
---|
140 | |
---|
141 | |
---|
142 | template <class C, class D> |
---|
143 | EnsembleBuilder<C, D>::~EnsembleBuilder(void) |
---|
144 | { |
---|
145 | for(size_t i=0; i<classifier_.size(); i++) |
---|
146 | delete classifier_[i]; |
---|
147 | delete subset_; |
---|
148 | } |
---|
149 | |
---|
150 | |
---|
151 | template <class C, class D> |
---|
152 | void EnsembleBuilder<C, D>::build(void) |
---|
153 | { |
---|
154 | for(u_long i=0; i<subset_->size();++i) { |
---|
155 | C* classifier = mother_.make_classifier(); |
---|
156 | classifier->train(subset_->training_data(i), |
---|
157 | subset_->training_target(i)); |
---|
158 | classifier_.push_back(classifier); |
---|
159 | } |
---|
160 | } |
---|
161 | |
---|
162 | |
---|
163 | template <class C, class D> |
---|
164 | const C& EnsembleBuilder<C, D>::classifier(size_t i) const |
---|
165 | { |
---|
166 | return *(classifier_[i]); |
---|
167 | } |
---|
168 | |
---|
169 | |
---|
170 | template <class C, class D> |
---|
171 | u_long EnsembleBuilder<C, D>::size(void) const |
---|
172 | { |
---|
173 | return classifier_.size(); |
---|
174 | } |
---|
175 | |
---|
176 | |
---|
177 | template <class C, class D> |
---|
178 | void EnsembleBuilder<C, D>::predict |
---|
179 | (const D& data, std::vector<std::vector<statistics::Averager> >& result) |
---|
180 | { |
---|
181 | result.clear(); |
---|
182 | result.reserve(subset_->target().nof_classes()); |
---|
183 | for(size_t i=0; i<subset_->target().nof_classes();i++) |
---|
184 | result.push_back(std::vector<statistics::Averager>(data.columns())); |
---|
185 | |
---|
186 | utility::Matrix prediction; |
---|
187 | |
---|
188 | for(u_long k=0;k<subset_->size();++k) { |
---|
189 | const D* sub_data = |
---|
190 | data.selected(subset_->training_features(k)); |
---|
191 | assert(sub_data); |
---|
192 | classifier(k).predict(*sub_data,prediction); |
---|
193 | delete sub_data; |
---|
194 | } |
---|
195 | |
---|
196 | for(size_t i=0; i<prediction.rows();i++) |
---|
197 | for(size_t j=0; j<prediction.columns();j++) |
---|
198 | result[i][j].add(prediction(i,j)); |
---|
199 | } |
---|
200 | |
---|
201 | |
---|
202 | template <class C, class D> |
---|
203 | const std::vector<std::vector<statistics::Averager> >& |
---|
204 | EnsembleBuilder<C, D>::validate(void) |
---|
205 | { |
---|
206 | validation_result_.clear(); |
---|
207 | |
---|
208 | validation_result_.reserve(subset_->target().nof_classes()); |
---|
209 | for(size_t i=0; i<subset_->target().nof_classes();i++) |
---|
210 | validation_result_.push_back(std::vector<statistics::Averager>(subset_->target().size())); |
---|
211 | |
---|
212 | utility::Matrix prediction; |
---|
213 | for(u_long k=0;k<subset_->size();k++) { |
---|
214 | classifier(k).predict(subset_->validation_data(k),prediction); |
---|
215 | |
---|
216 | // map results to indices of samples in training + validation data set |
---|
217 | for(size_t i=0; i<prediction.rows();i++) |
---|
218 | for(size_t j=0; j<prediction.columns();j++) { |
---|
219 | validation_result_[i][subset_->validation_index(k)[j]]. |
---|
220 | add(prediction(i,j)); |
---|
221 | } |
---|
222 | } |
---|
223 | return validation_result_; |
---|
224 | } |
---|
225 | |
---|
226 | }}} // of namespace classifier, yat, and theplu |
---|
227 | |
---|
228 | #endif |
---|