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: CrossEntropyVariable.cc 6351 2006-10-25 19:05:45Z chapados $ 00040 * This file is part of the PLearn library. 00041 ******************************************************* */ 00042 00043 #include "CrossEntropyVariable.h" 00044 00045 namespace PLearn { 00046 using namespace std; 00047 00048 00049 00052 PLEARN_IMPLEMENT_OBJECT(CrossEntropyVariable, 00053 "cost = - sum_i {target_i * log(output_i) + (1-target_i) * log(1-output_i)}", 00054 "NO HELP"); 00055 00056 CrossEntropyVariable::CrossEntropyVariable(Variable* netout, Variable* target) 00057 : inherited(netout,target,1,1) 00058 { 00059 build_(); 00060 } 00061 00062 void 00063 CrossEntropyVariable::build() 00064 { 00065 inherited::build(); 00066 build_(); 00067 } 00068 00069 void 00070 CrossEntropyVariable::build_() 00071 { 00072 // input1 and input2 are (respectively) netout and target from constructor 00073 if (input1 && input2 && (input1->size() != input2->size())) 00074 PLERROR("In CrossEntropyVariable: netout and target must have the same size"); 00075 } 00076 00077 00078 void CrossEntropyVariable::recomputeSize(int& l, int& w) const 00079 { l=1, w=1; } 00080 00081 void CrossEntropyVariable::fprop() 00082 { 00083 real cost = 0.0; 00084 for (int i=0; i<input1->size(); i++) 00085 { 00086 real output = input1->valuedata[i]; 00087 real target = input2->valuedata[i]; 00088 PLASSERT( fast_exact_is_equal(target, 0) || 00089 fast_exact_is_equal(target, 1) ); 00090 PLASSERT( output >= 0 && output <= 1 ); 00091 if ((fast_exact_is_equal(output,0) && !fast_exact_is_equal(target,0)) 00092 || (fast_exact_is_equal(output,1) && !fast_exact_is_equal(target,1))) 00093 PLERROR("CrossEntropyVariable::fprop: model output is either exactly " 00094 "0.0 or 1.0; cannot compute cost function"); 00095 if (!fast_exact_is_equal(output,0) && !fast_exact_is_equal(output,1)) { 00096 cost += target*pl_log(output) + (1.0-target)*pl_log(1.0-output); 00097 } 00098 } 00099 valuedata[0] = -cost; 00100 } 00101 00102 void CrossEntropyVariable::bprop() 00103 { 00104 real gr = *gradientdata; 00105 for (int i=0; i<input1->size(); i++) 00106 { 00107 real output = input1->valuedata[i]; 00108 real target = input2->valuedata[i]; 00109 #ifdef BOUNDCHECK 00110 if (fast_exact_is_equal(output, target)) 00111 PLERROR("CrossEntropyVariable::bprop: model output is either exactly " 00112 "0.0 or 1.0; cannot compute bprop"); 00113 #endif 00114 input1->gradientdata[i] += gr*(-target/output + (1.0-target)/(1.0-output)); 00115 } 00116 } 00117 00118 } // end of namespace PLearn 00119 00120 00121 /* 00122 Local Variables: 00123 mode:c++ 00124 c-basic-offset:4 00125 c-file-style:"stroustrup" 00126 c-file-offsets:((innamespace . 0)(inline-open . 0)) 00127 indent-tabs-mode:nil 00128 fill-column:79 00129 End: 00130 */ 00131 // vim: filetype=cpp:expandtab:shiftwidth=4:tabstop=8:softtabstop=4:encoding=utf-8:textwidth=79 :