Detailed Description

A hybrid Bayes net is a collection of HybridConditionals, which can have discrete conditionals, hybrid Gaussian conditionals, or pure Gaussian conditionals.

Inheritance diagram for gtsam::HybridBayesNet:

Public Member Functions
Standard Constructors
	HybridBayesNet ()=default
	Construct empty Bayes net.
	HybridBayesNet (std::initializer_list< HybridConditional::shared_ptr > conditionals)
	Constructor that takes an initializer list of shared pointers.
Testable
void	print (const std::string &s="", const KeyFormatter &formatter=DefaultKeyFormatter) const override
	GTSAM-style printing.
bool	equals (const This &fg, double tol=1e-9) const
	GTSAM-style equals.
Standard Interface
void	push_back (std::shared_ptr< HybridConditional > conditional)
	Add a hybrid conditional using a shared_ptr.
void	push_back (HybridConditional &&conditional)
	Move a HybridConditional into a shared pointer and add.
template<class CONDITIONAL>
void	push_back (const std::shared_ptr< CONDITIONAL > &conditional)
	Add a conditional to the Bayes net.
template<class CONDITIONAL, class... Args>
void	emplace_shared (Args &&...args)
	Preferred: Emplace a conditional directly using arguments.
DiscreteBayesNet	discreteMarginal () const
	Get the discrete Bayes Net P(M).
GaussianBayesNet	choose (const DiscreteValues &assignment) const
	Get the Gaussian Bayes net P(X\|M=m) corresponding to a specific assignment m for the discrete variables M.
double	evaluate (const HybridValues &values) const
	Evaluate hybrid probability density for given HybridValues.
double	operator() (const HybridValues &values) const
	Evaluate hybrid probability density for given HybridValues, sugar.
DiscreteValues	mpe () const
	Compute the Most Probable Explanation (MPE) of the discrete variables.
HybridValues	optimize () const
	Solve the HybridBayesNet by first computing the MPE of all the discrete variables and then optimizing the continuous variables based on the MPE assignment.
VectorValues	optimize (const DiscreteValues &assignment) const
	Given the discrete assignment, return the optimized estimate for the selected Gaussian BayesNet.
HybridValues	sample (const HybridValues &given, std::mt19937_64 *rng=nullptr) const
	Sample from an incomplete BayesNet, given missing variables.
HybridValues	sample (std::mt19937_64 *rng=nullptr) const
	Sample using ancestral sampling.
HybridBayesNet	prune (size_t maxNrLeaves, const std::optional< double > &marginalThreshold={}, DiscreteValues *fixedValues=nullptr) const
	Prune the Bayes Net such that we have at most maxNrLeaves leaves.
AlgebraicDecisionTree< Key >	errorTree (const VectorValues &continuousValues) const
	Compute the negative log posterior log P'(M\|x) of all assignments up to a constant, returning the result as an algebraic decision tree.
double	negLogConstant (const std::optional< DiscreteValues > &discrete={}) const
	Get the negative log of the normalization constant corresponding to the joint density represented by this Bayes net.
AlgebraicDecisionTree< Key >	discretePosterior (const VectorValues &continuousValues) const
	Compute normalized posterior P(M\|X=x) and return as a tree.
HybridGaussianFactorGraph	toFactorGraph (const VectorValues &measurements) const
	Convert a hybrid Bayes net to a hybrid Gaussian factor graph by converting all conditionals with instantiated measurements into likelihood factors.
double	logProbability (const HybridValues &x) const
Public Member Functions inherited from gtsam::BayesNet< HybridConditional >
void	dot (std::ostream &os, const KeyFormatter &keyFormatter=DefaultKeyFormatter, const DotWriter &writer=DotWriter()) const
	Output to graphviz format, stream version.
std::string	dot (const KeyFormatter &keyFormatter=DefaultKeyFormatter, const DotWriter &writer=DotWriter()) const
	Output to graphviz format string.
void	saveGraph (const std::string &filename, const KeyFormatter &keyFormatter=DefaultKeyFormatter, const DotWriter &writer=DotWriter()) const
	output to file with graphviz format.
double	logProbability (const HybridValues &x) const
double	evaluate (const HybridValues &c) const
virtual	~FactorGraph ()=default
	Default destructor Public and virtual so boost serialization can call it.
template<class DERIVEDFACTOR, typename = IsDerived<DERIVEDFACTOR>>
	FactorGraph (std::initializer_list< std::shared_ptr< DERIVEDFACTOR > > sharedFactors)
	Constructor that takes an initializer list of shared pointers.
void	reserve (size_t size)
	Reserve space for the specified number of factors if you know in advance how many there will be (works like FastVector::reserve).
template<class DERIVEDFACTOR>
IsDerived< DERIVEDFACTOR >	push_back (std::shared_ptr< DERIVEDFACTOR > factor)
	Add a factor directly using a shared_ptr.
template<class DERIVEDFACTOR, class... Args>
IsDerived< DERIVEDFACTOR >	emplace_shared (Args &&... args)
	Emplace a shared pointer to factor of given type.
template<class DERIVEDFACTOR>
IsDerived< DERIVEDFACTOR >	push_back (const DERIVEDFACTOR &factor)
	Add a factor by value, will be copy-constructed (use push_back with a shared_ptr to avoid the copy).
template<class DERIVEDFACTOR>
IsDerived< DERIVEDFACTOR >	add (std::shared_ptr< DERIVEDFACTOR > factor)
	add is a synonym for push_back.
template<class DERIVEDFACTOR>
std::enable_if< std::is_base_of< FactorType, DERIVEDFACTOR >::value, This >::type &	operator+= (std::shared_ptr< DERIVEDFACTOR > factor)
	Append factor to factor graph.
template<class DERIVEDFACTOR>
std::enable_if< std::is_base_of< FactorType, DERIVEDFACTOR >::value, This >::type &	operator, (std::shared_ptr< DERIVEDFACTOR > factor)
	Overload comma operator to allow for append chaining.
template<typename ITERATOR>
HasDerivedElementType< ITERATOR >	push_back (ITERATOR firstFactor, ITERATOR lastFactor)
	Push back many factors with an iterator over shared_ptr (factors are not copied).
template<typename ITERATOR>
HasDerivedValueType< ITERATOR >	push_back (ITERATOR firstFactor, ITERATOR lastFactor)
	Push back many factors with an iterator (factors are copied).
template<typename CONTAINER>
HasDerivedElementType< CONTAINER >	push_back (const CONTAINER &container)
	Push back many factors as shared_ptr's in a container (factors are not copied).
template<typename CONTAINER>
HasDerivedValueType< CONTAINER >	push_back (const CONTAINER &container)
	Push back non-pointer objects in a container (factors are copied).
template<class FACTOR_OR_CONTAINER>
void	add (const FACTOR_OR_CONTAINER &factorOrContainer)
	Add a factor or container of factors, including STL collections, BayesTrees, etc.
template<class FACTOR_OR_CONTAINER>
This &	operator+= (const FACTOR_OR_CONTAINER &factorOrContainer)
	Add a factor or container of factors, including STL collections, BayesTrees, etc.
template<class CLIQUE>
std::enable_if< std::is_base_of< This, typenameCLIQUE::FactorGraphType >::value >::type	push_back (const BayesTree< CLIQUE > &bayesTree)
	Push back a BayesTree as a collection of factors.
template<typename CONTAINER, typename = HasDerivedElementType<CONTAINER>>
FactorIndices	add_factors (const CONTAINER &factors, bool useEmptySlots=false)
	Add new factors to a factor graph and returns a list of new factor indices, optionally finding and reusing empty factor slots.
bool	equals (const This &fg, double tol=1e-9) const
	Check equality up to tolerance.
size_t	size () const
	return the number of factors (including any null factors set by remove() ).
bool	empty () const
	Check if the graph is empty (null factors set by remove() will cause this to return false).
const sharedFactor	at (size_t i) const
	Get a specific factor by index (this checks array bounds and may throw an exception, as opposed to operator[] which does not).
sharedFactor &	at (size_t i)
	Get a specific factor by index (this checks array bounds and may throw an exception, as opposed to operator[] which does not).
template<typename F>
std::shared_ptr< F >	at (size_t i)
	Get a specific factor by index and typecast to factor type F (this checks array bounds and may throw an exception, as opposed to operator[] which does not).
template<typename F>
const std::shared_ptr< F >	at (size_t i) const
	Const version of templated at method.
const sharedFactor	operator[] (size_t i) const
	Get a specific factor by index (this does not check array bounds, as opposed to at() which does).
sharedFactor &	operator[] (size_t i)
	Get a specific factor by index (this does not check array bounds, as opposed to at() which does).
const_iterator	begin () const
	Iterator to beginning of factors.
const_iterator	end () const
	Iterator to end of factors.
sharedFactor	front () const
	Get the first factor.
sharedFactor	back () const
	Get the last factor.
double	error (const HybridValues &values) const
	Add error for all factors.
iterator	begin ()
	non-const STL-style begin()
iterator	end ()
	non-const STL-style end()
virtual void	resize (size_t size)
	Directly resize the number of factors in the graph.
void	remove (size_t i)
	delete factor without re-arranging indexes by inserting a nullptr pointer
void	replace (size_t index, sharedFactor factor)
	replace a factor by index
iterator	erase (iterator item)
	Erase factor and rearrange other factors to take up the empty space.
iterator	erase (iterator first, iterator last)
	Erase factors and rearrange other factors to take up the empty space.
void	dot (std::ostream &os, const KeyFormatter &keyFormatter=DefaultKeyFormatter, const DotWriter &writer=DotWriter()) const
	Output to graphviz format, stream version.
std::string	dot (const KeyFormatter &keyFormatter=DefaultKeyFormatter, const DotWriter &writer=DotWriter()) const
	Output to graphviz format string.
void	saveGraph (const std::string &filename, const KeyFormatter &keyFormatter=DefaultKeyFormatter, const DotWriter &writer=DotWriter()) const
	output to file with graphviz format.
size_t	nrFactors () const
	return the number of non-null factors
KeySet	keys () const
	Potentially slow function to return all keys involved, sorted, as a set.
KeyVector	keyVector () const
	Potentially slow function to return all keys involved, sorted, as a vector.
bool	exists (size_t idx) const
	MATLAB interface utility: Checks whether a factor index idx exists in the graph and is a live pointer.

Public Types
using	Base = BayesNet<HybridConditional>
using	This = HybridBayesNet
using	ConditionalType = HybridConditional
using	shared_ptr = std::shared_ptr<HybridBayesNet>
using	sharedConditional = std::shared_ptr<ConditionalType>
Public Types inherited from gtsam::BayesNet< HybridConditional >
typedef std::shared_ptr< HybridConditional >	sharedConditional
	A shared pointer to a conditional.
Public Types inherited from gtsam::FactorGraph< FACTOR >
typedef FACTOR	FactorType
	factor type
typedef std::shared_ptr< FACTOR >	sharedFactor
	Shared pointer to a factor.
typedef sharedFactor	value_type
typedef FastVector< sharedFactor >::iterator	iterator
typedef FastVector< sharedFactor >::const_iterator	const_iterator

Additional Inherited Members
Protected Member Functions inherited from gtsam::BayesNet< HybridConditional >
	BayesNet ()
	Default constructor as an empty BayesNet.
	BayesNet (ITERATOR firstConditional, ITERATOR lastConditional)
	Construct from iterator over conditionals.
	BayesNet (std::initializer_list< sharedConditional > conditionals)
	Constructor that takes an initializer list of shared pointers.
Protected Member Functions inherited from gtsam::FactorGraph< FACTOR >
bool	isEqual (const FactorGraph &other) const
	Check exact equality of the factor pointers. Useful for derived ==.
	FactorGraph ()
	Default constructor.
template<typename ITERATOR>
	FactorGraph (ITERATOR firstFactor, ITERATOR lastFactor)
	Constructor from iterator over factors (shared_ptr or plain objects).
template<class CONTAINER>
	FactorGraph (const CONTAINER &factors)
	Construct from container of factors (shared_ptr or plain objects).
Protected Attributes inherited from gtsam::FactorGraph< FACTOR >
FastVector< sharedFactor >	factors_
	concept check, makes sure FACTOR defines print and equals

Member Function Documentation

◆ choose()

GaussianBayesNet gtsam::HybridBayesNet::choose ( const DiscreteValues & assignment ) const

Get the Gaussian Bayes net P(X|M=m) corresponding to a specific assignment m for the discrete variables M.

As the hybrid Bayes net defines P(X,M) = P(X|M) P(M), this method returns the posterior p(X|M=m).

Parameters

assignment The discrete value assignment for the discrete keys.

Returns: Gaussian posterior P(X|M=m) as a GaussianBayesNet.

◆ discreteMarginal()

DiscreteBayesNet gtsam::HybridBayesNet::discreteMarginal ( ) const

Get the discrete Bayes Net P(M).

As the hybrid Bayes net defines P(X,M) = P(X|M) P(M), this method returns the marginal distribution on the discrete variables.

Returns: discrete marginal as a DiscreteBayesNet.

◆ discretePosterior()

AlgebraicDecisionTree< Key > gtsam::HybridBayesNet::discretePosterior ( const VectorValues & continuousValues ) const

Compute normalized posterior P(M|X=x) and return as a tree.

Note: Not a DiscreteConditional as the cardinalities of the DiscreteKeys, which we would need, are hard to recover.

Parameters

continuousValues Continuous values x to condition P(M|X=x) on.

Returns: AlgebraicDecisionTree<Key>

◆ emplace_shared()

template<class CONDITIONAL, class... Args>

void gtsam::HybridBayesNet::emplace_shared ( Args &&... args )

inline

Preferred: Emplace a conditional directly using arguments.

Examples: hbn.emplace_shared<HybridGaussianConditional>(...))); hbn.emplace_shared<GaussianConditional>(...))); hbn.emplace_shared<DiscreteConditional>(...)));

◆ errorTree()

AlgebraicDecisionTree< Key > gtsam::HybridBayesNet::errorTree ( const VectorValues & continuousValues ) const

Compute the negative log posterior log P'(M|x) of all assignments up to a constant, returning the result as an algebraic decision tree.

Note: The joint P(X,M) is p(X|M) P(M) Then the posterior on M given X=x is is P(M|x) = p(x|M) P(M) / p(x). Ideally we want log P(M|x) = log p(x|M) + log P(M) - log p(x), but unfortunately log p(x) is expensive, so we compute the log of the unnormalized posterior log P'(M|x) = log p(x|M) + log P(M)

Parameters

continuousValues Continuous values x at which to compute log P'(M|x)

Returns: AlgebraicDecisionTree<Key>

◆ mpe()

DiscreteValues gtsam::HybridBayesNet::mpe ( ) const

Compute the Most Probable Explanation (MPE) of the discrete variables.

Returns: DiscreteValues

◆ negLogConstant()

double gtsam::HybridBayesNet::negLogConstant ( const std::optional< DiscreteValues > & discrete = {} ) const

Get the negative log of the normalization constant corresponding to the joint density represented by this Bayes net.

Optionally index by Discrete.

Parameters

Discrete Optional DiscreteValues

Returns: double

◆ optimize() [1/2]

HybridValues gtsam::HybridBayesNet::optimize ( ) const

Solve the HybridBayesNet by first computing the MPE of all the discrete variables and then optimizing the continuous variables based on the MPE assignment.

Returns: HybridValues

◆ optimize() [2/2]

VectorValues gtsam::HybridBayesNet::optimize ( const DiscreteValues & assignment ) const

Given the discrete assignment, return the optimized estimate for the selected Gaussian BayesNet.

Parameters

assignment An assignment of discrete values.

Returns: Values

◆ print()

void gtsam::HybridBayesNet::print	(	const std::string &	s = "",
		const KeyFormatter &	formatter = DefaultKeyFormatter ) const

overridevirtual

GTSAM-style printing.

Reimplemented from gtsam::BayesNet< HybridConditional >.

◆ prune()

HybridBayesNet gtsam::HybridBayesNet::prune	(	size_t	maxNrLeaves,
		const std::optional< double > &	marginalThreshold = {},
		DiscreteValues *	fixedValues = nullptr ) const

Prune the Bayes Net such that we have at most maxNrLeaves leaves.

Parameters

maxNrLeaves	Continuous values at which to compute the error.
marginalThreshold	The threshold to check the mode marginals against.
fixedValues	The fixed values resulting from dead mode removal.

Note: If marginal greater than this threshold, the mode gets assigned that value and is considered "dead" for hybrid elimination. The mode can then be removed since it only has a single possible assignment.

Returns: A pruned HybridBayesNet

◆ push_back() [1/3]

template<class CONDITIONAL>

void gtsam::HybridBayesNet::push_back ( const std::shared_ptr< CONDITIONAL > & conditional )

inline

Add a conditional to the Bayes net.

Implicitly convert to a HybridConditional.

E.g. hbn.push_back(std::make_shared<DiscreteConditional>(m, "1/1"));

Template Parameters

CONDITIONAL Type of conditional. This is shared_ptr version.

Parameters

conditional The conditional as a shared pointer.

◆ push_back() [2/3]

void gtsam::HybridBayesNet::push_back ( HybridConditional && conditional )

inline

Move a HybridConditional into a shared pointer and add.

Example: HybridGaussianConditional conditional(...); hbn.push_back(conditional); // loses the original conditional

◆ push_back() [3/3]

void gtsam::HybridBayesNet::push_back ( std::shared_ptr< HybridConditional > conditional )

inline

Add a hybrid conditional using a shared_ptr.

This is the "native" push back, as this class stores hybrid conditionals.

◆ sample() [1/2]

HybridValues gtsam::HybridBayesNet::sample	(	const HybridValues &	given,
		std::mt19937_64 *	rng = nullptr ) const

Sample from an incomplete BayesNet, given missing variables.

Example: std::mt19937_64 rng(42); VectorValues given = ...; auto sample = bn.sample(given, &rng);

Parameters

given	Values of missing variables.
rng	The optional pseudo-random number generator.

Returns: HybridValues

◆ sample() [2/2]

HybridValues gtsam::HybridBayesNet::sample ( std::mt19937_64 * rng = nullptr ) const

Sample using ancestral sampling.

Example: std::mt19937_64 rng(42); auto sample = bn.sample(&rng);

Parameters

rng	The optional pseudo-random number generator.

Returns: HybridValues

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

/tmp/gtsam-4.3.0-doxygen.rsXPUS/source/gtsam/hybrid/HybridBayesNet.h
/tmp/gtsam-4.3.0-doxygen.rsXPUS/source/gtsam/hybrid/HybridBayesNet.cpp

Detailed Description

Public Member Functions

Public Types

Additional Inherited Members

Member Function Documentation

◆ choose()

◆ discreteMarginal()

◆ discretePosterior()

◆ emplace_shared()

◆ errorTree()

◆ mpe()

◆ negLogConstant()

◆ optimize() [1/2]

◆ optimize() [2/2]

◆ print()

◆ prune()

◆ push_back() [1/3]

◆ push_back() [2/3]

◆ push_back() [3/3]

◆ sample() [1/2]

◆ sample() [2/2]