Improved HAK (added 'maxtime' property)
[libdai.git] / src / clustergraph.cpp
1 /* This file is part of libDAI - http://www.libdai.org/
2 *
3 * libDAI is licensed under the terms of the GNU General Public License version
4 * 2, or (at your option) any later version. libDAI is distributed without any
5 * warranty. See the file COPYING for more details.
6 *
7 * Copyright (C) 2006-2010 Joris Mooij [joris dot mooij at libdai dot org]
8 * Copyright (C) 2006-2007 Radboud University Nijmegen, The Netherlands
9 */
10
11
12 #include <set>
13 #include <vector>
14 #include <iostream>
15 #include <dai/varset.h>
16 #include <dai/clustergraph.h>
17
18
19 namespace dai {
20
21
22 using namespace std;
23
24
25 ClusterGraph::ClusterGraph( const std::vector<VarSet> & cls ) : _G(), _vars(), _clusters() {
26 // construct vars, clusters and edge list
27 vector<Edge> edges;
28 foreach( const VarSet &cl, cls ) {
29 if( find( clusters().begin(), clusters().end(), cl ) == clusters().end() ) {
30 // add cluster
31 size_t n2 = nrClusters();
32 _clusters.push_back( cl );
33 for( VarSet::const_iterator n = cl.begin(); n != cl.end(); n++ ) {
34 size_t n1 = find( vars().begin(), vars().end(), *n ) - vars().begin();
35 if( n1 == nrVars() )
36 // add variable
37 _vars.push_back( *n );
38 edges.push_back( Edge( n1, n2 ) );
39 }
40 } // disregard duplicate clusters
41 }
42
43 // Create bipartite graph
44 _G.construct( nrVars(), nrClusters(), edges.begin(), edges.end() );
45 }
46
47
48 size_t sequentialVariableElimination::operator()( const ClusterGraph &cl, const std::set<size_t> &/*remainingVars*/ ) {
49 return cl.findVar( seq.at(i++) );
50 }
51
52
53 size_t greedyVariableElimination::operator()( const ClusterGraph &cl, const std::set<size_t> &remainingVars ) {
54 set<size_t>::const_iterator lowest = remainingVars.end();
55 size_t lowest_cost = -1UL;
56 for( set<size_t>::const_iterator i = remainingVars.begin(); i != remainingVars.end(); i++ ) {
57 size_t cost = heuristic( cl, *i );
58 if( lowest == remainingVars.end() || lowest_cost > cost ) {
59 lowest = i;
60 lowest_cost = cost;
61 }
62 }
63 return *lowest;
64 }
65
66
67 size_t eliminationCost_MinNeighbors( const ClusterGraph &cl, size_t i ) {
68 return cl.bipGraph().delta1( i ).size();
69 }
70
71
72 size_t eliminationCost_MinWeight( const ClusterGraph &cl, size_t i ) {
73 SmallSet<size_t> id_n = cl.bipGraph().delta1( i );
74
75 size_t cost = 1;
76 for( SmallSet<size_t>::const_iterator it = id_n.begin(); it != id_n.end(); it++ )
77 cost *= cl.vars()[*it].states();
78
79 return cost;
80 }
81
82
83 size_t eliminationCost_MinFill( const ClusterGraph &cl, size_t i ) {
84 SmallSet<size_t> id_n = cl.bipGraph().delta1( i );
85
86 size_t cost = 0;
87 // for each unordered pair {i1,i2} adjacent to n
88 for( SmallSet<size_t>::const_iterator it1 = id_n.begin(); it1 != id_n.end(); it1++ )
89 for( SmallSet<size_t>::const_iterator it2 = it1; it2 != id_n.end(); it2++ )
90 if( it1 != it2 ) {
91 // if i1 and i2 are not adjacent, eliminating n would make them adjacent
92 if( !cl.adj(*it1, *it2) )
93 cost++;
94 }
95
96 return cost;
97 }
98
99
100 size_t eliminationCost_WeightedMinFill( const ClusterGraph &cl, size_t i ) {
101 SmallSet<size_t> id_n = cl.bipGraph().delta1( i );
102
103 size_t cost = 0;
104 // for each unordered pair {i1,i2} adjacent to n
105 for( SmallSet<size_t>::const_iterator it1 = id_n.begin(); it1 != id_n.end(); it1++ )
106 for( SmallSet<size_t>::const_iterator it2 = it1; it2 != id_n.end(); it2++ )
107 if( it1 != it2 ) {
108 // if i1 and i2 are not adjacent, eliminating n would make them adjacent
109 if( !cl.adj(*it1, *it2) )
110 cost += cl.vars()[*it1].states() * cl.vars()[*it2].states();
111 }
112
113 return cost;
114 }
115
116
117 } // end of namespace dai