include/dai/hak.h

   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 #ifndef __defined_libdai_hak_h
  23 #define __defined_libdai_hak_h
  24
  25
  26 #include <string>
  27 #include <dai/daialg.h>
  28 #include <dai/regiongraph.h>
  29 #include <dai/enum.h>
  30 #include <dai/properties.h>
  31
  32
  33 namespace dai {
  34
  35
  36 /// HAK provides an implementation of the single and double-loop algorithms by Heskes, Albers and Kappen
  37 class HAK : public DAIAlgRG {
  38     protected:
  39         std::vector<Factor>                _Qa;
  40         std::vector<Factor>                _Qb;
  41         std::vector<std::vector<Factor> >  _muab;
  42         std::vector<std::vector<Factor> >  _muba;
  43         /// Maximum difference encountered so far
  44         double _maxdiff;
  45         /// Number of iterations needed
  46         size_t _iters;
  47
  48     public:
  49         struct Properties {
  50             size_t verbose;
  51             size_t maxiter;
  52             double tol;
  53             double damping;
  54             DAI_ENUM(ClustersType,MIN,DELTA,LOOP)
  55             ClustersType clusters;
  56             bool doubleloop;
  57             size_t loopdepth;
  58         } props;
  59         /// Name of this inference method
  60         static const char *Name;
  61
  62     public:
  63         /// Default constructor
  64         HAK() : DAIAlgRG(), _Qa(), _Qb(), _muab(), _muba(), _maxdiff(0.0), _iters(0U), props() {}
  65
  66         /// Construct from FactorGraph fg and PropertySet opts
  67         HAK( const FactorGraph &fg, const PropertySet &opts );
  68
  69         /// Construct from RegionGraph rg and PropertySet opts
  70         HAK( const RegionGraph &rg, const PropertySet &opts );
  71
  72         /// Copy constructor
  73         HAK( const HAK &x ) : DAIAlgRG(x), _Qa(x._Qa), _Qb(x._Qb), _muab(x._muab), _muba(x._muba), _maxdiff(x._maxdiff), _iters(x._iters), props(x.props) {}
  74
  75         /// Clone *this (virtual copy constructor)
  76         virtual HAK* clone() const { return new HAK(*this); }
  77
  78         /// Create (virtual default constructor)
  79         virtual HAK* create() const { return new HAK(); }
  80
  81         /// Assignment operator
  82         HAK& operator=( const HAK &x ) {
  83             if( this != &x ) {
  84                 DAIAlgRG::operator=( x );
  85                 _Qa      = x._Qa;
  86                 _Qb      = x._Qb;
  87                 _muab    = x._muab;
  88                 _muba    = x._muba;
  89                 _maxdiff = x._maxdiff;
  90                 _iters   = x._iters;
  91                 props    = x.props;
  92             }
  93             return *this;
  94         }
  95
  96         /// Identifies itself for logging purposes
  97         virtual std::string identify() const;
  98
  99         /// Get single node belief
 100         virtual Factor belief( const Var &n ) const;
 101
 102         /// Get general belief
 103         virtual Factor belief( const VarSet &ns ) const;
 104
 105         /// Get all beliefs
 106         virtual std::vector<Factor> beliefs() const;
 107
 108         /// Get log partition sum
 109         virtual Real logZ() const;
 110
 111         /// Clear messages and beliefs
 112         virtual void init();
 113
 114         /// Clear messages and beliefs corresponding to the nodes in ns
 115         virtual void init( const VarSet &ns );
 116
 117         /// The actual approximate inference algorithm
 118         virtual double run();
 119
 120         /// Return maximum difference between single node beliefs in the last pass
 121         virtual double maxDiff() const { return _maxdiff; }
 122
 123         /// Return number of passes over the factorgraph
 124         virtual size_t Iterations() const { return _iters; }
 125
 126
 127         Factor & muab( size_t alpha, size_t _beta ) { return _muab[alpha][_beta]; }
 128         Factor & muba( size_t alpha, size_t _beta ) { return _muba[alpha][_beta]; }
 129         const Factor& Qa( size_t alpha ) const { return _Qa[alpha]; };
 130         const Factor& Qb( size_t beta ) const { return _Qb[beta]; };
 131
 132         double doGBP();
 133         double doDoubleLoop();
 134
 135         void setProperties( const PropertySet &opts );
 136         PropertySet getProperties() const;
 137         std::string printProperties() const;
 138
 139     private:
 140         void constructMessages();
 141         void findLoopClusters( const FactorGraph &fg, std::set<VarSet> &allcl, VarSet newcl, const Var & root, size_t length, VarSet vars );
 142 };
 143
 144
 145 } // end of namespace dai
 146
 147
 148 #endif