libDAI
Todo List
File alldai.h

Replace VarSets by SmallSet<size_t> where appropriate, in order to minimize the use of FactorGraph::findVar().

Improve SWIG interfaces and merge their build process with the main build process

File bbp.h
Clean up code
File bp.h
Consider using a priority_queue for maximum residual schedule
File clustergraph.h
The "MinFill" and "WeightedMinFill" variable elimination heuristics seem designed for Markov graphs; add similar heuristics which are designed for factor graphs.
Member dai::BP::setProperties (const PropertySet &opts)
Make DAI_BP_FAST a compile-time choice, as it is a memory/speed tradeoff
Member dai::CBP::getInfAlg ()
At present, CBP::getInfAlg() only returns a BP instance; it should be possible to select other inference algorithms via a property
Class dai::ClusterGraph
Remove the _vars and _clusters variables and use only the graph and a contextual factor graph.
Class dai::CobwebGraph
Implement unit test for Cobwebgraph
Member dai::CobwebGraph::Connection::newmsg
Remove CobwebGraph::Connection::newmsg
Class dai::FactorGraph
Write a method that applies evidence (should we represent evidence as a map<Var,size_t> or as a map<size_t,size_t>?)
Class dai::FBP
Add nice way to set weights
Author
Frederik Eaton
Class dai::TFactor< T >

Define a better fileformat for .fg files (maybe using XML)?

Add support for sparse factors.

Class dai::TRWBP
Merge code of FBP and TRWBP
Member dai::TRWBP::nrtrees
See if there is a way to wrap TRWBP::nrtrees in a props struct together with the other properties currently in TRWBP::props (without copying al lot of BP code literally)
File emalg.h
Implement parameter estimation for undirected models / factor graphs.
File glc.h
Fix the init of GLC
File hak.h

Use ClusterGraph instead of a vector<VarSet> for speed.

Optimize this code for large factor graphs.

Implement GBP parent-child algorithm.

File treeep.h
Clean up the TreeEP code (exploiting that a large part of the code is just a special case of JTree).
File weightedgraph.h
Improve general support for graphs and trees (in particular, a good tree implementation is needed).