PLearn 0.1
|
00001 // -*- C++ -*- 00002 00003 // PLearn (A C++ Machine Learning Library) 00004 // Copyright (C) 1998 Pascal Vincent 00005 // Copyright (C) 1999-2001 Pascal Vincent, Yoshua Bengio, Rejean Ducharme and University of Montreal 00006 // Copyright (C) 2002 Pascal Vincent, Julien Keable, Xavier Saint-Mleux 00007 // 00008 // Redistribution and use in source and binary forms, with or without 00009 // modification, are permitted provided that the following conditions are met: 00010 // 00011 // 1. Redistributions of source code must retain the above copyright 00012 // notice, this list of conditions and the following disclaimer. 00013 // 00014 // 2. Redistributions in binary form must reproduce the above copyright 00015 // notice, this list of conditions and the following disclaimer in the 00016 // documentation and/or other materials provided with the distribution. 00017 // 00018 // 3. The name of the authors may not be used to endorse or promote 00019 // products derived from this software without specific prior written 00020 // permission. 00021 // 00022 // THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND ANY EXPRESS OR 00023 // IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 00024 // OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN 00025 // NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 00026 // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED 00027 // TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 00028 // PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 00029 // LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 00030 // NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 00031 // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 00032 // 00033 // This file is part of the PLearn library. For more information on the PLearn 00034 // library, go to the PLearn Web site at www.plearn.org 00035 00036 00037 /* ******************************************************* 00038 * $Id: GeneralizedOneHotVMatrix.cc 7479 2007-06-01 00:43:49Z larocheh $ 00039 ******************************************************* */ 00040 00041 #include "GeneralizedOneHotVMatrix.h" 00042 00043 namespace PLearn { 00044 using namespace std; 00045 00046 00049 PLEARN_IMPLEMENT_OBJECT(GeneralizedOneHotVMatrix, 00050 "VMatrix that maps many columns to a one-hot vector", 00051 "This VMat is a generalization of OneHotVMatrix where many columns (given\n" 00052 "by the Vec index) are mapped, instead of just the last one."); 00053 00054 GeneralizedOneHotVMatrix::GeneralizedOneHotVMatrix(bool call_build_) 00055 : inherited(call_build_) 00056 { 00057 if( call_build_ ) 00058 build_(); 00059 } 00060 00061 GeneralizedOneHotVMatrix::GeneralizedOneHotVMatrix(VMat the_source, 00062 Vec the_index, 00063 Vec the_nclasses, 00064 Vec the_cold_value, 00065 Vec the_hot_value, 00066 bool call_build_) 00067 : inherited(the_source, 00068 the_source->length(), 00069 the_source->width() 00070 + (int)sum(the_nclasses)-the_nclasses.length(), 00071 call_build_), 00072 index(the_index), nclasses(the_nclasses), 00073 cold_value(the_cold_value), hot_value(the_hot_value) 00074 { 00075 if (min(index)<0 || max(index)>source->length()-1) 00076 PLERROR("In GeneralizedOneHotVMatrix: all values of index must be in" 00077 " range [0,%d]", 00078 source->length()-1); 00079 00080 if (index.length()!=nclasses.length() || 00081 cold_value.length()!=hot_value.length() || 00082 index.length()!=hot_value.length()) 00083 PLERROR("In GeneralizedOneHotVMatrix: index, nclasses, cold_value\n" 00084 "and hot_value must have same length.\n"); 00085 00086 if (call_build_) 00087 build_(); 00088 } 00089 00090 void GeneralizedOneHotVMatrix::build() 00091 { 00092 inherited::build(); 00093 build_(); 00094 } 00095 00096 void GeneralizedOneHotVMatrix::build_() 00097 { 00098 length_ = source->length(); 00099 width_ = source->width() + (int)sum(nclasses) - nclasses.length(); 00100 00101 TVec<string> fieldnames = source->fieldNames(); 00102 TVec<string> extended_fieldnames( width() ); 00103 int efn_pos = 0; 00104 for( int j=0 ; j<fieldnames.length() ; j++ ) 00105 { 00106 const int index_pos = vec_find(index, (real)j); 00107 if( index_pos == -1 ) 00108 extended_fieldnames[efn_pos++] = fieldnames[j]; 00109 else 00110 { 00111 const int nb_class = (int)nclasses[index_pos]; 00112 string name = fieldnames[j]; 00113 for( int k=0 ; k<nb_class ; k++ ) 00114 extended_fieldnames[efn_pos++] = name + "_" + tostring(k); 00115 00116 } 00117 } 00118 declareFieldNames( extended_fieldnames ); 00119 setMetaInfoFromSource(); 00120 } 00121 00122 void GeneralizedOneHotVMatrix::declareOptions(OptionList &ol) 00123 { 00124 declareOption(ol, "distr", &GeneralizedOneHotVMatrix::source, 00125 (OptionBase::learntoption | OptionBase::nosave), 00126 "DEPRECATED - use 'source' instead."); 00127 declareOption(ol, "index", &GeneralizedOneHotVMatrix::index, 00128 OptionBase::buildoption, "Columns to map to one-hot vector representation."); 00129 declareOption(ol, "nclasses", &GeneralizedOneHotVMatrix::nclasses, 00130 OptionBase::buildoption, "Size of the one-hot vector for each columns to map."); 00131 declareOption(ol, "cold_value", &GeneralizedOneHotVMatrix::cold_value, 00132 OptionBase::buildoption, "Cold values for all columns to map."); 00133 declareOption(ol, "hot_value", &GeneralizedOneHotVMatrix::hot_value, 00134 OptionBase::buildoption, "Hot values for all columns to map."); 00135 inherited::declareOptions(ol); 00136 } 00137 00138 void GeneralizedOneHotVMatrix::getNewRow(int i, const Vec& v) const 00139 { 00140 #ifdef BOUNDCHECK 00141 if(i<0 || i>=length()) 00142 PLERROR("In OneHotVMatrix::getNewRow OUT OF BOUNDS"); 00143 if(v.length()!=width()) 00144 PLERROR("In GeneralizedOneHotVMatrix::getNewRow v.length() must be\n" 00145 "equal to the VMat's width.\n"); 00146 #endif 00147 00148 Vec input(source->width()); 00149 source->getRow(i, input); 00150 int v_pos = 0; 00151 for (int j=0; j<input.length(); j++) { 00152 const int index_pos = vec_find(index, (real)j); 00153 if (index_pos == -1) 00154 v[v_pos++] = input[j]; 00155 else { 00156 const int nb_class = (int)nclasses[index_pos]; 00157 Vec target = v.subVec(v_pos, nb_class); 00158 const real cold = cold_value[index_pos]; 00159 const real hot = hot_value[index_pos]; 00160 const real classnum = source->get(i,j); // Maybe not a desired behavior 00161 if(is_missing(classnum)) target.fill(cold); 00162 else fill_one_hot(target, (int) classnum, cold, hot); 00163 v_pos += nb_class; 00164 } 00165 } 00166 } 00167 00169 // makeDeepCopyFromShallowCopy // 00171 void GeneralizedOneHotVMatrix::makeDeepCopyFromShallowCopy(CopiesMap& copies) 00172 { 00173 inherited::makeDeepCopyFromShallowCopy(copies); 00174 00175 // ### Call deepCopyField on all "pointer-like" fields. 00176 // ### that you wish to be deepCopied rather than. 00177 // ### shallow-copied. 00178 // ### ex: 00179 // deepCopyField(trainvec, copies); 00180 00181 deepCopyField(index, copies); 00182 deepCopyField(nclasses, copies); 00183 deepCopyField(cold_value, copies); 00184 deepCopyField(hot_value, copies); 00185 } 00186 00187 00188 00189 } // end of namespace PLearn 00190 00191 00192 /* 00193 Local Variables: 00194 mode:c++ 00195 c-basic-offset:4 00196 c-file-style:"stroustrup" 00197 c-file-offsets:((innamespace . 0)(inline-open . 0)) 00198 indent-tabs-mode:nil 00199 fill-column:79 00200 End: 00201 */ 00202 // vim: filetype=cpp:expandtab:shiftwidth=4:tabstop=8:softtabstop=4:encoding=utf-8:textwidth=79 :