BNCore Bayesian Network Inference

Graph model, dense factors, junction tree compilation and batched exact inference. More...

Files
file	manager.hpp
	bncore::DiscretizationManager, a helper that refines the state list of a Graph variable by splitting one state (bin) into two, up to a configured limit.
file	dense_tensor.hpp
	bncore::DenseTensor, a row-major multi-dimensional array of doubles that either owns its storage or borrows a caller buffer.
file	factor.hpp
	bncore::Factor, a table of doubles over a list of variables, with the product and sum-out operations of discrete inference.
file	graph.hpp
	bncore::Graph, the discrete Bayesian network model: variables, their states, directed edges and conditional probability tables.
file	node.hpp
	Basic BNCore types: the NodeId variable handle and the VariableMetadata record that Graph stores for each variable.
file	compiler.hpp
	bncore::JunctionTreeCompiler, which turns a Graph into the JunctionTree that the inference engine runs on.
file	engine.hpp
	bncore::BatchExecutionEngine, the query interface of BNCore: exact posterior marginals and MAP states for batches of evidence rows on a compiled JunctionTree.
file	junction_tree.hpp
	Junction tree data structure of BNCore: Clique, Separator and the JunctionTree container.
file	workspace.hpp
	bncore::BatchWorkspace, the message passing state of junction tree inference, and the precomputed index tables it runs on (CollectOp, DistributeOp, AssembleOp, CliqueMargInfo).
file	bump_allocator.hpp
	bncore::BumpAllocator, a fixed-capacity arena of doubles for temporary factor storage.

Detailed Description

Graph model, dense factors, junction tree compilation and batched exact inference.