gtsam  4.1.0
gtsam
gtsam::BayesTreeCliqueBase< DERIVED, FACTORGRAPH > Class Template Reference

Detailed Description

template<class DERIVED, class FACTORGRAPH>
class gtsam::BayesTreeCliqueBase< DERIVED, FACTORGRAPH >

This is the base class for BayesTree cliques.

The default and standard derived type is BayesTreeClique, but some algorithms, like iSAM2, use a different clique type in order to store extra data along with the clique.

This class is templated on the derived class (i.e. the curiously recursive template pattern). The advantage of this over using virtual classes is that it avoids the need for casting to get the derived type. This is possible because all cliques in a BayesTree are the same type - if they were not then we'd need a virtual class.

Template Parameters
DERIVEDThe derived clique type.
CONDITIONALThe conditional type.

Standard Constructors

 BayesTreeCliqueBase ()
 Default constructor.
 
 BayesTreeCliqueBase (const sharedConditional &conditional)
 Construct from a conditional, leaving parent and child pointers uninitialized.
 
 BayesTreeCliqueBase (const BayesTreeCliqueBase &c)
 Shallow copy constructor.
 
BayesTreeCliqueBaseoperator= (const BayesTreeCliqueBase &c)
 Shallow copy assignment constructor.
 

Testable

bool equals (const DERIVED &other, double tol=1e-9) const
 check equality
 
virtual void print (const std::string &s="", const KeyFormatter &keyFormatter=DefaultKeyFormatter) const
 print this node
 

Standard Interface

const sharedConditional & conditional () const
 Access the conditional.
 
bool isRoot () const
 is this the root of a Bayes tree ?
 
size_t treeSize () const
 The size of subtree rooted at this clique, i.e., nr of Cliques.
 
size_t numCachedSeparatorMarginals () const
 Collect number of cliques with cached separator marginals.
 
derived_ptr parent () const
 return a shared_ptr to the parent clique
 
int problemSize () const
 Problem size (used for parallel traversal)
 

Advanced Interface

class BayesTree< DerivedType >
 
class boost::serialization::access
 Serialization function.
 
KeyVector separator_setminus_B (const derived_ptr &B) const
 Calculate set \( S \setminus B \) for shortcut calculations.
 
KeyVector shortcut_indices (const derived_ptr &B, const FactorGraphType &p_Cp_B) const
 Determine variable indices to keep in recursive separator shortcut calculation The factor graph p_Cp_B has keys from the parent clique Cp and from B. More...
 
void deleteCachedShortcutsNonRecursive ()
 Non-recursive delete cached shortcuts and marginals - internal only.
 
BayesNetType shortcut (const derived_ptr &root, Eliminate function=EliminationTraitsType::DefaultEliminate) const
 return the conditional P(S|Root) on the separator given the root
 
FactorGraphType separatorMarginal (Eliminate function=EliminationTraitsType::DefaultEliminate) const
 return the marginal P(S) on the separator
 
FactorGraphType marginal2 (Eliminate function=EliminationTraitsType::DefaultEliminate) const
 return the marginal P(C) of the clique, using marginal caching
 
void deleteCachedShortcuts ()
 This deletes the cached shortcuts of all cliques (subtree) below this clique. More...
 
const boost::optional< FactorGraphType > & cachedSeparatorMarginal () const
 

Public Member Functions

void setEliminationResult (const typename FactorGraphType::EliminationResult &eliminationResult)
 Fill the elimination result produced during elimination. More...
 

Public Attributes

sharedConditional conditional_
 
derived_weak_ptr parent_
 
FastVector< derived_ptr > children
 
int problemSize_
 
bool is_root = false
 

Public Types

typedef FACTORGRAPH FactorGraphType
 
typedef EliminationTraitsType::BayesNetType BayesNetType
 
typedef BayesNetType::ConditionalType ConditionalType
 
typedef boost::shared_ptr< ConditionalType > sharedConditional
 
typedef FactorGraphType::FactorType FactorType
 
typedef FactorGraphType::Eliminate Eliminate
 

Protected Attributes

boost::optional< FactorGraphType > cachedSeparatorMarginal_
 This stores the Cached separator marginal P(S)
 
std::mutex cachedSeparatorMarginalMutex_
 This protects Cached seperator marginal P(S) from concurrent read/writes as many the functions which access it are const (hence the mutable) leading to the false impression that these const functions are thread-safe which is not true due to these mutable values. More...
 

Member Function Documentation

◆ deleteCachedShortcuts()

template<class DERIVED , class FACTORGRAPH >
void gtsam::BayesTreeCliqueBase< DERIVED, FACTORGRAPH >::deleteCachedShortcuts

This deletes the cached shortcuts of all cliques (subtree) below this clique.

This is performed when the bayes tree is modified.

◆ setEliminationResult()

template<class DERIVED , class FACTORGRAPH >
void gtsam::BayesTreeCliqueBase< DERIVED, FACTORGRAPH >::setEliminationResult ( const typename FactorGraphType::EliminationResult &  eliminationResult)

Fill the elimination result produced during elimination.

Here this just stores the conditional and ignores the remaining factor, but this is overridden in ISAM2Clique to also cache the remaining factor.

◆ shortcut_indices()

template<class DERIVED , class FACTORGRAPH >
KeyVector gtsam::BayesTreeCliqueBase< DERIVED, FACTORGRAPH >::shortcut_indices ( const derived_ptr &  B,
const FactorGraphType &  p_Cp_B 
) const
protected

Determine variable indices to keep in recursive separator shortcut calculation The factor graph p_Cp_B has keys from the parent clique Cp and from B.

But we only keep the variables not in S union B.

Member Data Documentation

◆ cachedSeparatorMarginalMutex_

template<class DERIVED , class FACTORGRAPH >
std::mutex gtsam::BayesTreeCliqueBase< DERIVED, FACTORGRAPH >::cachedSeparatorMarginalMutex_
mutableprotected

This protects Cached seperator marginal P(S) from concurrent read/writes as many the functions which access it are const (hence the mutable) leading to the false impression that these const functions are thread-safe which is not true due to these mutable values.

This is fixed by applying this mutex.


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