ErrorFunction.h
Go to the documentation of this file.
1 /*!
2  *
3  *
4  * \brief error function for supervised learning
5  *
6  *
7  *
8  * \author T.Voss, T. Glasmachers, O.Krause
9  * \date 2010-2011
10  *
11  *
12  * \par Copyright 1995-2017 Shark Development Team
13  *
14  * <BR><HR>
15  * This file is part of Shark.
16  * <http://shark-ml.org/>
17  *
18  * Shark is free software: you can redistribute it and/or modify
19  * it under the terms of the GNU Lesser General Public License as published
20  * by the Free Software Foundation, either version 3 of the License, or
21  * (at your option) any later version.
22  *
23  * Shark is distributed in the hope that it will be useful,
24  * but WITHOUT ANY WARRANTY; without even the implied warranty of
25  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26  * GNU Lesser General Public License for more details.
27  *
28  * You should have received a copy of the GNU Lesser General Public License
29  * along with Shark. If not, see <http://www.gnu.org/licenses/>.
30  *
31  */
32 #ifndef SHARK_OBJECTIVEFUNCTIONS_ERRORFUNCTION_H
33 #define SHARK_OBJECTIVEFUNCTIONS_ERRORFUNCTION_H
34 
35 
36 #include <shark/Models/AbstractModel.h>
37 #include <shark/ObjectiveFunctions/Loss/AbstractLoss.h>
38 #include <shark/ObjectiveFunctions/AbstractObjectiveFunction.h>
39 #include <shark/Data/Dataset.h>
40 #include <shark/Data/WeightedDataset.h>
41 #include "Impl/FunctionWrapperBase.h"
42 
43 #include <boost/scoped_ptr.hpp>
44 
45 namespace shark{
46 
47 ///
48 /// \brief Objective function for supervised learning
49 ///
50 /// \par
51 /// An ErrorFunction object is an objective function for
52 /// learning the parameters of a model from data by means
53 /// of minimization of a cost function. The value of the
54 /// objective function is the cost of the model predictions
55 /// on the training data, given the targets.
56 /// \par
57 /// It supports mini-batch learning using an optional fourth argument to
58 /// The constructor. With mini-batch learning enabled, each iteration a random
59 /// batch is taken from the dataset. Thus the size of the minibatch is the size of the batches in
60 /// the datasets. Normalization ensures that batches of different sizes have approximately the same
61 /// magnitude of error and derivative.
62 ///
63 ///\par
64 /// It automatically infers the input und label type from the given dataset and the output type
65 /// of the model in the constructor and ensures that Model and loss match. Thus the user does
66 /// not need to provide the types as template parameters.
67 class ErrorFunction : public SingleObjectiveFunction
68 {
69 public:
70  template<class InputType, class LabelType, class OutputType>
71  ErrorFunction(
72  LabeledData<InputType, LabelType> const& dataset,
73  AbstractModel<InputType,OutputType>* model,
74  AbstractLoss<LabelType, OutputType>* loss,
75  bool useMiniBatches = false
76  );
77  template<class InputType, class LabelType, class OutputType>
78  ErrorFunction(
79  WeightedLabeledData<InputType, LabelType> const& dataset,
80  AbstractModel<InputType,OutputType>* model,
81  AbstractLoss<LabelType, OutputType>* loss
82  );
83  ErrorFunction(const ErrorFunction& op);
84  ErrorFunction& operator=(const ErrorFunction& op);
85 
86  std::string name() const
87  { return "ErrorFunction"; }
88 
89  void setRegularizer(double factor, SingleObjectiveFunction* regularizer){
90  m_regularizer = regularizer;
91  m_regularizationStrength = factor;
92  }
93 
94  SearchPointType proposeStartingPoint()const {
95  return mp_wrapper -> proposeStartingPoint();
96  }
97  std::size_t numberOfVariables()const{
98  return mp_wrapper -> numberOfVariables();
99  }
100 
101  void init(){
102  mp_wrapper->setRng(this->mep_rng);
103  mp_wrapper-> init();
104  }
105 
106  double eval(RealVector const& input) const;
107  ResultType evalDerivative( const SearchPointType & input, FirstOrderDerivative & derivative ) const;
108 
109  friend void swap(ErrorFunction& op1, ErrorFunction& op2);
110 
111 private:
112  boost::scoped_ptr<detail::FunctionWrapperBase > mp_wrapper;
113  SingleObjectiveFunction* m_regularizer;
114  double m_regularizationStrength;
115 };
116 
117 }
118 #include "Impl/ErrorFunction.inl"
119 #endif