-- Distributed under the Boost Software License, Version 1.0. -- (See accompanying file LICENSE_1_0.txt or copy at -- http://www.boost.org/LICENSE_1_0.txt)
cuthill_mckee_ordering
| Graphs: | undirected |
|---|---|
| Properties: | color, degree |
| Complexity: | time: O(m log(m)|V|) where m = max { degree(v) | v in V } |
(1)
template <class IncidenceGraph, class OutputIterator,
class ColorMap, class DegreeMap>
OutputIterator
cuthill_mckee_ordering(const IncidenceGraph& g,
typename graph_traits<Graph>::vertex_descriptor s,
OutputIterator inverse_permutation,
ColorMap color, DegreeMap degree)
(2)
template <class VertexListGraph, class OutputIterator>
OutputIterator
cuthill_mckee_ordering(const VertexIndexMap& G, OutputIterator inverse_permutation);
template <class VertexListGraph, class OutputIterator, class VertexIndexMap>
OutputIterator
cuthill_mckee_ordering(const VertexIndexMap& G, OutputIterator inverse_permutation,
VertexIndexMap index_map);
template <class VertexListGraph, class OutputIterator,
class ColorMap, class DegreeMap>
OutputIterator
cuthill_mckee_ordering(const VertexListGraph& G, OutputIterator inverse_permutation,
ColorMap color, DegreeMap degree)
(3)
template <class IncidenceGraph, class OutputIterator,
class ColorMap, class DegreeMap>
OutputIterator
cuthill_mckee_ordering(const IncidenceGraph& g,
std::deque< typename
graph_traits<Graph>::vertex_descriptor > vertex_queue,
OutputIterator inverse_permutation,
ColorMap color, DegreeMap degree)
The goal of the Cuthill-Mckee (and reverse Cuthill-Mckee) ordering
algorithm[14, 43, 44, 45 ] is to reduce the bandwidth of a graph by reordering the
indices assigned to each vertex. The Cuthill-Mckee ordering algorithm
works by a local minimization of the i-th bandwidths. The vertices are
basically assigned a breadth-first search order, except that at each
step, the adjacent vertices are placed in the queue in order of
increasing degree.
Version 1 of the algorithm lets the user choose the ``starting vertex'', version 2 finds a good starting vertex using the pseudo-peripheral pair heuristic (among each component), while version 3 contains the starting nodes for each vertex in the deque. The choice of the ``starting vertex'' can have a significant effect on the quality of the ordering. For versions 2 and 3, find_starting_vertex will be called for each component in the graph, increasing run time significantly.
The output of the algorithm are the vertices in the new ordering. Depending on what kind of output iterator you use, you can get either the Cuthill-Mckee ordering or the reverse Cuthill-McKee ordering. For example, if you store the output into a vector using the vector's reverse iterator, then you get the reverse Cuthill-McKee ordering.
std::vector<vertex_descriptor> inv_perm(num_vertices(G)); cuthill_mckee_ordering(G, inv_perm.rbegin(), ...);
Either way, storing the output into a vector gives you the permutation from the new ordering to the old ordering.
inv_perm[new_index[u]] == u
Often times, it is the opposite permutation that you want, the permutation from the old index to the new index. This can easily be computed in the following way.
for (size_type i = 0; i != inv_perm.size(); ++i)
perm[old_index[inv_perm[i]]] = i;
| Copyright © 2000-2001 | Jeremy Siek, Indiana University (jsiek@osl.iu.edu) |