PLearn 0.1
|
00001 // -*- C++ -*- 00002 00003 // PLearn (A C++ Machine Learning Library) 00004 // Copyright (C) 1998 Pascal Vincent 00005 // Copyright (C) 1999-2002 Pascal Vincent, Yoshua Bengio, Rejean Ducharme and University of Montreal 00006 // Copyright (C) 2001-2002 Nicolas Chapados, Ichiro Takeuchi, Jean-Sebastien Senecal 00007 // Copyright (C) 2002 Xiangdong Wang, Christian Dorion 00008 00009 // Redistribution and use in source and binary forms, with or without 00010 // modification, are permitted provided that the following conditions are met: 00011 // 00012 // 1. Redistributions of source code must retain the above copyright 00013 // notice, this list of conditions and the following disclaimer. 00014 // 00015 // 2. Redistributions in binary form must reproduce the above copyright 00016 // notice, this list of conditions and the following disclaimer in the 00017 // documentation and/or other materials provided with the distribution. 00018 // 00019 // 3. The name of the authors may not be used to endorse or promote 00020 // products derived from this software without specific prior written 00021 // permission. 00022 // 00023 // THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND ANY EXPRESS OR 00024 // IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 00025 // OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN 00026 // NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 00027 // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED 00028 // TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 00029 // PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 00030 // LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 00031 // NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 00032 // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 00033 // 00034 // This file is part of the PLearn library. For more information on the PLearn 00035 // library, go to the PLearn Web site at www.plearn.org 00036 00037 00038 /* ******************************************************* 00039 * $Id: BiasWeightAffineTransformVariable.cc 3994 2005-08-25 13:35:03Z chapados $ 00040 * This file is part of the PLearn library. 00041 ******************************************************* */ 00042 00043 #include "BiasWeightAffineTransformVariable.h" 00044 #include <plearn/math/TMat_maths.h> 00045 00046 namespace PLearn { 00047 using namespace std; 00048 00049 00050 PLEARN_IMPLEMENT_OBJECT(BiasWeightAffineTransformVariable, 00051 "Affine transformation of a vector variable, from a weight and bias variable.", 00052 "NO HELP"); 00053 00054 BiasWeightAffineTransformVariable::BiasWeightAffineTransformVariable( 00055 const VarArray& the_varray, 00056 bool the_transpose_weights) 00057 : inherited(the_varray, 00058 (the_varray[0]->size() == 1) ? 00059 the_varray[2]->size() : 00060 (the_varray[0]->isRowVec() ? 1 : the_varray[2]->size()), 00061 (the_varray[0]->size() == 1) ? 00062 1 : (the_varray[0]->isRowVec() ? the_varray[2]->size() : 1)), 00063 transpose_weights(the_transpose_weights) 00064 { 00065 build_(); 00066 } 00067 00068 void 00069 BiasWeightAffineTransformVariable::makeDeepCopyFromShallowCopy(CopiesMap& copies) 00070 { 00071 inherited::makeDeepCopyFromShallowCopy(copies); 00072 varDeepCopyField(vec, copies); 00073 varDeepCopyField(weights, copies); 00074 varDeepCopyField(bias, copies); 00075 } 00076 00077 00078 void 00079 BiasWeightAffineTransformVariable::build() 00080 { 00081 inherited::build(); 00082 build_(); 00083 } 00084 00085 void 00086 BiasWeightAffineTransformVariable::build_() 00087 { 00088 if(varray.length() != 3) 00089 PLERROR("In BiasWeightAffineTransformVariable():build_(): needs an input" 00090 " vector, a weight matrix and a bias vector"); 00091 00092 vec = varray[0]; 00093 weights = varray[1]; 00094 bias = varray[2]; 00095 00096 if (!vec->isVec()) 00097 PLERROR("In BiasWeightAffineTransformVariable: expecting a vector Var" 00098 " for the input vector"); 00099 if(vec->size() != (!transpose_weights ? weights->length() : weights->width())) 00100 PLERROR("In BiasWeightAffineTransformVariable: weight matrix has" 00101 " incompatible size"); 00102 } 00103 00104 void BiasWeightAffineTransformVariable::recomputeSize(int& l, int& w) const 00105 { 00106 if (vec && weights) { 00107 l = (vec->isRowVec() && vec->size() != 1) ? 1 : bias->size(); 00108 w = vec->isColumnVec() ? 1 : bias->size(); 00109 } else 00110 l = w = 0; 00111 } 00112 00113 00114 void BiasWeightAffineTransformVariable::fprop() 00115 { 00116 value << bias->value; 00117 00118 // Seems contradictive, but we are considering x.T versus x.T', 00119 // not T.x versus T'.x, for consistency with AffineTransformVariable! 00120 if(transpose_weights) 00121 productAcc(value, weights->matValue, vec->value); 00122 else 00123 transposeProductAcc(value, weights->matValue, vec->value); 00124 } 00125 00126 00127 void BiasWeightAffineTransformVariable::bprop() 00128 { 00129 bias->gradient += gradient; 00130 if(!vec->dont_bprop_here) 00131 { 00132 if(transpose_weights) 00133 transposeProductAcc(vec->gradient, weights->matValue, gradient); 00134 else 00135 productAcc(vec->gradient, weights->matValue, gradient); 00136 } 00137 00138 if(transpose_weights) 00139 externalProductAcc(weights->matGradient, gradient, vec->value); 00140 else 00141 externalProductAcc(weights->matGradient, vec->value, gradient); 00142 } 00143 00144 00145 void BiasWeightAffineTransformVariable::symbolicBprop() 00146 { 00147 PLERROR("BiasWeightAffineTransformVariable::symbolicBprop() not implemented"); 00148 } 00149 00150 } // end of namespace PLearn 00151 00152 00153 /* 00154 Local Variables: 00155 mode:c++ 00156 c-basic-offset:4 00157 c-file-style:"stroustrup" 00158 c-file-offsets:((innamespace . 0)(inline-open . 0)) 00159 indent-tabs-mode:nil 00160 fill-column:79 00161 End: 00162 */ 00163 // vim: filetype=cpp:expandtab:shiftwidth=4:tabstop=8:softtabstop=4:encoding=utf-8:textwidth=79 :