1 - Idea: a FactorGraph and a RegionGraph are often equipped with
2 extra properties for nodes and edges. The code to initialize those
3 is often quite similar; maybe this can be abstracted to classes
4 like ExtFactorGraph and ExtRegionGraph (Ext stands for Extended), e.g.
5 template <typename NodeProperties, typename EdgeProperties>
6 class ExtFactorGraph : public FactorGraph {
8 std::vector<NodeProperties> nodeProps;
9 std::vector<std::vector<EdgeProperties> > edgeProps;
12 A disadvantage of this approach may be worse cachability.
13 - http://www.boost.org/development/requirements.html#Design_and_Programming
14 - Would it be a good idea to cache second-order neighborhoods (delta's) in BipGraph?
15 - Would it be a good idea to add the variable label -> index hashmap in FactorGraph,
16 to replace the linear searches that are performed every time for findVar()?
17 No, a better idea is to avoid calls to findVar() as much as possible.
18 - Can the FactorGraph constructors be optimized further?
21 - Test Visual C++ compatibility
22 - Figure out which libraries are required and document in README
23 boost headers, boost::program_options library, boost::graph library,
24 boost::math library (under Windows)
26 FILES IN SVN HEAD THAT ARE NO LONGER RELEVANT FOR GIT MASTER