PLearn 0.1
AdaBoost.h
Go to the documentation of this file.
00001 // -*- C++ -*-
00002 
00003 // AdaBoost.h
00004 //
00005 // Copyright (C) 2003  Pascal Vincent 
00006 // 
00007 // Redistribution and use in source and binary forms, with or without
00008 // modification, are permitted provided that the following conditions are met:
00009 // 
00010 //  1. Redistributions of source code must retain the above copyright
00011 //     notice, this list of conditions and the following disclaimer.
00012 // 
00013 //  2. Redistributions in binary form must reproduce the above copyright
00014 //     notice, this list of conditions and the following disclaimer in the
00015 //     documentation and/or other materials provided with the distribution.
00016 // 
00017 //  3. The name of the authors may not be used to endorse or promote
00018 //     products derived from this software without specific prior written
00019 //     permission.
00020 // 
00021 // THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND ANY EXPRESS OR
00022 // IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
00023 // OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN
00024 // NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
00025 // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
00026 // TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
00027 // PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
00028 // LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
00029 // NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
00030 // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
00031 // 
00032 // This file is part of the PLearn library. For more information on the PLearn
00033 // library, go to the PLearn Web site at www.plearn.org
00034 
00035 /* *******************************************************      
00036  * $Id: AdaBoost.h 10097 2009-04-07 15:58:04Z nouiz $
00037  ******************************************************* */
00038 
00039 // Authors: Yoshua Bengio
00040 
00043 #ifndef AdaBoost_INC
00044 #define AdaBoost_INC
00045 
00046 #include <plearn_learners/generic/PLearner.h>
00047 
00048 namespace PLearn {
00049 using namespace std;
00050 
00051 class AdaBoost: public PLearner
00052 {
00053     typedef PLearner inherited;
00054 
00056     mutable Vec weighted_costs;
00057     mutable Vec sum_weighted_costs;
00058     mutable Vec weak_learner_output;
00059 
00061     mutable TVec<VMat> saved_testset;
00062     mutable TVec<VMat> saved_testoutputs;
00063     mutable TVec<int>  saved_last_test_stages;
00064 
00065 protected:
00066     // average weighted error of each learner
00067     Vec learners_error;
00068     // weighing scheme over examples
00069     Vec example_weights;
00070 
00071     // *********************
00072     // * protected options *
00073     // *********************
00074 
00075     // saved options:
00076     // (unnormalized) weight associated to each weak learner
00077     Vec voting_weights;
00078     real sum_voting_weights; // = sum(voting_weights);
00079     real initial_sum_weights;
00080 
00082     TVec< PP<PLearner> > weak_learners;
00083 
00085     bool found_zero_error_weak_learner;
00086 
00087 public:
00088 
00089     // ************************
00090     // * public build options *
00091     // ************************
00092 
00097     PP<PLearner> weak_learner_template;
00098 
00099     // normally 0.5
00100     real target_error;
00101 
00102     // whether to give an expdir to the underlying weak learners
00103     bool provide_learner_expdir;
00104   
00105     // threshold on output of weak learner to decide if class 0 or class 1
00106     real output_threshold;
00107 
00108     // whether to compute training error during training
00109     bool compute_training_error;
00110 
00111     // use Pseudo-loss Adaboost
00112     bool pseudo_loss_adaboost;
00113 
00114     // use Confidence-rated adaboost
00115     bool conf_rated_adaboost;
00116 
00117     // use resampling (vs weighting) to train the underlying classifier
00118     bool weight_by_resampling;
00119 
00120     // stop if weak learner does not seem to help
00121     bool early_stopping;
00122 
00123     // save model after each stage into <expdir>/model.psave
00124     bool save_often;
00125 
00126     // Did we add the sub_learner_costs to our costs
00127     bool forward_sub_learner_test_costs;
00128 
00129     // Did we modif directly the train_set weights?
00130     bool modif_train_set_weights;
00131 
00132     // Did we save and reuse previous test result?
00133     // This is usefull to have a test time that is 
00134     // independent of the number of adaboost itaration
00135     bool reuse_test_results;
00136     // ****************
00137     // * Constructors *
00138     // ****************
00139 
00140     // Default constructor, make sure the implementation in the .cc
00141     // initializes all fields to reasonable default values.
00142     AdaBoost();
00143 
00144 
00145     // ******************
00146     // * PLearner methods *
00147     // ******************
00148 
00149 private: 
00151     void build_();
00152 
00153     void computeTrainingError(Vec input, Vec target);
00154 
00155     void computeOutput_(const Vec& input, Vec& output,
00156                         const int start=0, const real sum=0.) const;
00157 
00158 protected: 
00160     // (Please implement in .cc)
00161     static void declareOptions(OptionList& ol);
00162 
00163 public:
00164 
00165     // ************************
00166     // **** Object methods ****
00167     // ************************
00168 
00170     virtual void build();
00171 
00173     virtual void makeDeepCopyFromShallowCopy(CopiesMap& copies);
00174 
00175     // Declares other standard object methods
00176     //  If your class is not instantiatable (it has pure virtual methods)
00177     // you should replace this by PLEARN_DECLARE_ABSTRACT_OBJECT_METHODS 
00178     PLEARN_DECLARE_OBJECT(AdaBoost);
00179 
00180 
00181     // **************************
00182     // **** PLearner methods ****
00183     // **************************
00184 
00189     virtual int outputsize() const;
00190 
00191     virtual void finalize();
00192 
00195     virtual void forget();
00196 
00199     virtual void train();
00200     virtual void test(VMat testset, PP<VecStatsCollector> test_stats,
00201                       VMat testoutputs, VMat testcosts) const;
00202 
00204     virtual void computeOutput(const Vec& input, Vec& output) const{
00205         computeOutput_(input,output,0,0);}
00206     virtual void computeOutputAndCosts(const Vec& input, const Vec& target,
00207                                        Vec& output, Vec& costs) const;
00208 
00210     virtual void computeCostsFromOutputs(const Vec& input, const Vec& output, 
00211                                          const Vec& target, Vec& costs) const;
00212                                 
00213 
00215     virtual TVec<string> getTestCostNames() const;
00216 
00219     virtual TVec<string> getTrainCostNames() const;
00220     virtual void         setTrainingSet(VMat training_set, bool call_forget=true);
00221 
00222 
00223 };
00224 
00225 // Declares a few other classes and functions related to this class
00226 DECLARE_OBJECT_PTR(AdaBoost);
00227   
00228 } // end of namespace PLearn
00229 
00230 #endif
00231 
00232 
00233 /*
00234   Local Variables:
00235   mode:c++
00236   c-basic-offset:4
00237   c-file-style:"stroustrup"
00238   c-file-offsets:((innamespace . 0)(inline-open . 0))
00239   indent-tabs-mode:nil
00240   fill-column:79
00241   End:
00242 */
00243 // vim: filetype=cpp:expandtab:shiftwidth=4:tabstop=8:softtabstop=4:encoding=utf-8:textwidth=79 :
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Defines