src/mf.cpp

   1 /*  Copyright (C) 2006-2008  Joris Mooij  [j dot mooij at science dot ru dot nl]
   2     Radboud University Nijmegen, The Netherlands
   3
   4     This file is part of libDAI.
   5
   6     libDAI is free software; you can redistribute it and/or modify
   7     it under the terms of the GNU General Public License as published by
   8     the Free Software Foundation; either version 2 of the License, or
   9     (at your option) any later version.
  10
  11     libDAI is distributed in the hope that it will be useful,
  12     but WITHOUT ANY WARRANTY; without even the implied warranty of
  13     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14     GNU General Public License for more details.
  15
  16     You should have received a copy of the GNU General Public License
  17     along with libDAI; if not, write to the Free Software
  18     Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  19 */
  20
  21
  22 #include <iostream>
  23 #include <sstream>
  24 #include <map>
  25 #include <set>
  26 #include <dai/mf.h>
  27 #include <dai/diffs.h>
  28 #include <dai/util.h>
  29
  30
  31 namespace dai {
  32
  33
  34 using namespace std;
  35
  36
  37 const char *MF::Name = "MF";
  38
  39
  40 bool MF::checkProperties() {
  41     if( !HasProperty("tol") )
  42         return false;
  43     if (!HasProperty("maxiter") )
  44         return false;
  45     if (!HasProperty("verbose") )
  46         return false;
  47
  48     ConvertPropertyTo<double>("tol");
  49     ConvertPropertyTo<size_t>("maxiter");
  50     ConvertPropertyTo<size_t>("verbose");
  51
  52     return true;
  53 }
  54
  55
  56 void MF::create() {
  57     // clear beliefs
  58     _beliefs.clear();
  59     _beliefs.reserve( nrVars() );
  60
  61     // create beliefs
  62     for( size_t i = 0; i < nrVars(); ++i )
  63         _beliefs.push_back(Factor(var(i)));
  64 }
  65
  66
  67 string MF::identify() const {
  68     stringstream result (stringstream::out);
  69     result << Name << GetProperties();
  70     return result.str();
  71 }
  72
  73
  74 void MF::init() {
  75     assert( checkProperties() );
  76
  77     for( vector<Factor>::iterator qi = _beliefs.begin(); qi != _beliefs.end(); qi++ )
  78         qi->fill(1.0);
  79 }
  80
  81
  82 double MF::run() {
  83     double tic = toc();
  84
  85     if( Verbose() >= 1 )
  86         cout << "Starting " << identify() << "...";
  87
  88     size_t pass_size = _beliefs.size();
  89     Diffs diffs(pass_size * 3, 1.0);
  90
  91     size_t t=0;
  92     for( t=0; t < (MaxIter()*pass_size) && diffs.maxDiff() > Tol(); t++ ) {
  93         // choose random Var i
  94         size_t i = (size_t) (nrVars() * rnd_uniform());
  95
  96         Factor jan;
  97         Factor piet;
  98         foreach( const Neighbor &I, nbV(i) ) {
  99             Factor henk;
 100             foreach( const Neighbor &j, nbF(I) ) // for all j in I \ i
 101                 if( j != i )
 102                     henk *= _beliefs[j];
 103             piet = factor(I).log0();
 104             piet *= henk;
 105             piet = piet.part_sum(var(i));
 106             piet = piet.exp();
 107             jan *= piet;
 108         }
 109
 110         jan.normalize( _normtype );
 111
 112         if( jan.hasNaNs() ) {
 113             cout << "MF::run():  ERROR: jan has NaNs!" << endl;
 114             return NAN;
 115         }
 116
 117         diffs.push( dist( jan, _beliefs[i], Prob::DISTLINF ) );
 118
 119         _beliefs[i] = jan;
 120     }
 121
 122     updateMaxDiff( diffs.maxDiff() );
 123
 124     if( Verbose() >= 1 ) {
 125         if( diffs.maxDiff() > Tol() ) {
 126             if( Verbose() == 1 )
 127                 cout << endl;
 128             cout << "MF::run:  WARNING: not converged within " << MaxIter() << " passes (" << toc() - tic << " clocks)...final maxdiff:" << diffs.maxDiff() << endl;
 129         } else {
 130             if( Verbose() >= 2 )
 131                 cout << "MF::run:  ";
 132             cout << "converged in " << t / pass_size << " passes (" << toc() - tic << " clocks)." << endl;
 133         }
 134     }
 135
 136     return diffs.maxDiff();
 137 }
 138
 139
 140 Factor MF::beliefV (size_t i) const {
 141     Factor piet;
 142     piet = _beliefs[i];
 143     piet.normalize( Prob::NORMPROB );
 144     return(piet);
 145 }
 146
 147
 148 Factor MF::belief (const VarSet &ns) const {
 149     if( ns.size() == 1 )
 150         return belief( *(ns.begin()) );
 151     else {
 152         assert( ns.size() == 1 );
 153         return Factor();
 154     }
 155 }
 156
 157
 158 Factor MF::belief (const Var &n) const {
 159     return( beliefV( findVar( n ) ) );
 160 }
 161
 162
 163 vector<Factor> MF::beliefs() const {
 164     vector<Factor> result;
 165     for( size_t i = 0; i < nrVars(); i++ )
 166         result.push_back( beliefV(i) );
 167     return result;
 168 }
 169
 170
 171 Complex MF::logZ() const {
 172     Complex sum = 0.0;
 173
 174     for(size_t i=0; i < nrVars(); i++ )
 175         sum -= beliefV(i).entropy();
 176     for(size_t I=0; I < nrFactors(); I++ ) {
 177         Factor henk;
 178         foreach( const Neighbor &j, nbF(I) )  // for all j in I
 179             henk *= _beliefs[j];
 180         henk.normalize( Prob::NORMPROB );
 181         Factor piet;
 182         piet = factor(I).log0();
 183         piet *= henk;
 184         sum -= Complex( piet.totalSum() );
 185     }
 186
 187     return -sum;
 188 }
 189
 190
 191 void MF::init( const VarSet &ns ) {
 192     for( size_t i = 0; i < nrVars(); i++ ) {
 193         if( ns.contains(var(i) ) )
 194             _beliefs[i].fill( 1.0 );
 195     }
 196 }
 197
 198
 199 } // end of namespace dai