Darwin  1.10(beta)
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Groups Pages
Public Member Functions | Static Public Attributes | List of all members
drwnDualDecompositionInference Class Reference

Implements dual decomposition MAP inference (see Komodakis and Paragios, CVPR 2009 and works cited therein). Each factor is treated as a separate slave. More...

Inheritance diagram for drwnDualDecompositionInference:
drwnMAPInference

Public Member Functions

 drwnDualDecompositionInference (const drwnFactorGraph &graph)
 
std::pair< double, double > inference (drwnFullAssignment &mapAssignment)
 Run inference (or resume for iterative algorithms). Algorithms may initialize from mapAssignment if not empty. Returns an upper and lower bound (if available) of the minimum energy. The upper bound is the same as the energy of the best solution found (i.e., same as graph.getEnergy(mapAssignment)). More...
 
- Public Member Functions inherited from drwnMAPInference
 drwnMAPInference (const drwnFactorGraph &graph)
 
 drwnMAPInference (const drwnMAPInference &inf)
 
virtual void clear ()
 Clear internally cached data (e.g., computation graph)
 

Static Public Attributes

static double INITIAL_ALPHA = 0.5
 initial gradient step size
 
static bool USE_MIN_MARGINALS = false
 use min-marginals for subgradients
 

Additional Inherited Members

- Protected Attributes inherited from drwnMAPInference
const drwnFactorGraph_graph
 reference to initial clique potentials
 

Detailed Description

Implements dual decomposition MAP inference (see Komodakis and Paragios, CVPR 2009 and works cited therein). Each factor is treated as a separate slave.

Member Function Documentation

pair< double, double > drwnDualDecompositionInference::inference ( drwnFullAssignment mapAssignment)
virtual

Run inference (or resume for iterative algorithms). Algorithms may initialize from mapAssignment if not empty. Returns an upper and lower bound (if available) of the minimum energy. The upper bound is the same as the energy of the best solution found (i.e., same as graph.getEnergy(mapAssignment)).

Todo:
change for min-marginal subgradient

Implements drwnMAPInference.


The documentation for this class was generated from the following files: