gtsam::Sampler Class Reference

Detailed Description

Sampling structure that keeps internal random number generators for diagonal distributions specified by NoiseModel.

basic functionality
Vector	sample () const
	sample from distribution
template<typename T>
T	perturb (const T &value) const
	Perturb a value by sampling in its tangent space and applying retract.
static Vector	sampleDiagonal (const Vector &sigmas, std::mt19937_64 *rng)
	sample with given random number generator

Public Member Functions
constructors
	Sampler (const noiseModel::Diagonal::shared_ptr &model, uint_fast64_t seed=42u)
	Create a sampler for the distribution specified by a diagonal NoiseModel with a manually specified seed.
	Sampler (const noiseModel::Diagonal::shared_ptr &model, std::mt19937_64 &rng)
	Create a sampler that draws from a caller-supplied RNG (stateful).
	Sampler (const Vector &sigmas, uint_fast64_t seed=42u)
	Create a sampler for a distribution specified by a vector of sigmas directly.
	Sampler (const Vector &sigmas, std::mt19937_64 &rng)
	Create a sampler for sigmas that draws from a caller-supplied RNG (stateful).
access functions
size_t	dim () const
Vector	sigmas () const
const noiseModel::Diagonal::shared_ptr &	model () const

Public Types
typedef std::shared_ptr< Sampler >	shared_ptr

Protected Member Functions
Vector	sampleDiagonal (const Vector &sigmas) const
	Given sigmas for a diagonal model, returns a sample.

Protected Attributes
noiseModel::Diagonal::shared_ptr	model_
	noiseModel created at generation
std::mt19937_64	generator_
	generator
std::mt19937_64 *	externalGenerator_ = nullptr
	Non-owning optional external generator. If non-null, sampling uses this.

Constructor & Destructor Documentation

Sampler() [1/4]

gtsam::Sampler::Sampler ( const noiseModel::Diagonal::shared_ptr & model, uint_fast64_t seed = 42u )

explicit

Create a sampler for the distribution specified by a diagonal NoiseModel with a manually specified seed.

This constructor is convenient for deterministic, throw-away sampling. If you need stateful sampling across calls or across multiple Sampler instances, prefer the RNG-based constructor and manage the RNG yourself.

NOTE: do not use zero as a seed, it will break the generator.

Sampler() [2/4]

gtsam::Sampler::Sampler ( const noiseModel::Diagonal::shared_ptr & model, std::mt19937_64 & rng )

explicit

Create a sampler that draws from a caller-supplied RNG (stateful).

The RNG is non-owning and must outlive this Sampler.

Sampler() [3/4]

gtsam::Sampler::Sampler ( const Vector & sigmas, uint_fast64_t seed = 42u )

explicit

Create a sampler for a distribution specified by a vector of sigmas directly.

This constructor is convenient for deterministic, throw-away sampling. If you need stateful sampling across calls or across multiple Sampler instances, prefer the RNG-based constructor and manage the RNG yourself.

NOTE: do not use zero as a seed, it will break the generator.

Sampler() [4/4]

gtsam::Sampler::Sampler ( const Vector & sigmas, std::mt19937_64 & rng )

explicit

Create a sampler for sigmas that draws from a caller-supplied RNG (stateful).

The RNG is non-owning and must outlive this Sampler.

Member Function Documentation

perturb()

template<typename T>

T gtsam::Sampler::perturb ( const T & value ) const

inline

Perturb a value by sampling in its tangent space and applying retract.

The supplied noise model must match the dimensionality expected by T.

sampleDiagonal()

Vector gtsam::Sampler::sampleDiagonal ( const Vector & sigmas ) const

protected

Given sigmas for a diagonal model, returns a sample.

Uses external RNG if available.

---

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

• /tmp/gtsam-4.3.0-doxygen.rsXPUS/source/gtsam/linear/Sampler.h
• /tmp/gtsam-4.3.0-doxygen.rsXPUS/source/gtsam/linear/Sampler.cpp