infrared.infrared.MultiDimensionalBoltzmannSampler Class Reference

Multi-dimensional Boltzmann sampler. More...

Inheritance diagram for infrared.infrared.MultiDimensionalBoltzmannSampler:

Collaboration diagram for infrared.infrared.MultiDimensionalBoltzmannSampler:

Public Member Functions
	__init__ (self, model, td_factory=TreeDecompositionFactory(), lazy=True)
	Construct with model and optional td_factory.
	is_good_sample (self, features, values)
	whether the sample is of good quality
	set_target (self, target, tolerance, featureid)
	Set target of a feature.
	targeted_samples (self)
	Generator of targeted samples.
	targeted_sample (self)
Public Member Functions inherited from infrared.infrared.BoltzmannSampler
	__init__ (self, model, td_factory=TreeDecompositionFactory(), lazy=True)
	Construct.
	sample (self)
	Generate sample.
	resample (self, variables, assignment)
	Generate sample.
	samples (self)
	Sample generator.
	gen_cluster_tree (self)
	Suitable cluster tree.
Public Member Functions inherited from infrared.infrared.EngineBase
	__init__ (self, model, td_factory=TreeDecompositionFactory(), lazy=True)
	Construct with model and optional td_factory.
	requires_reinitialization (self)
	setup_engine (self, *skip_ct=False)
	Sets up the constraint model / cluster tree sampling engine.
	evaluate (self)
	Evaluates the cluster tree.
	is_consistent (self)
	plot_td (self, filename, to="pdf")
	Plot the tree decomposition to pdf file filename write to filename to target format, support conversion to "pdf" or "png".
	treewidth (self)
	Get tree width.
	gen_cluster_tree (self)
	Generate the populated cluster tree td tree decomposition Returns cluster tree.

Static Public Member Functions
	rmsd (means, features)

Public Attributes
	samples_per_round
	tweak_factor
	cooling
	verbose
	callback

Protected Attributes
	_targeted_samples
Protected Attributes inherited from infrared.infrared.BoltzmannSampler
	_model
Protected Attributes inherited from infrared.infrared.EngineBase
	_model
	_td_factory
	_td
	_ct

Additional Inherited Members
Properties inherited from infrared.infrared.EngineBase
	model = property
	td = property
	ct = property

Detailed Description

Multi-dimensional Boltzmann sampler.

Constructor & Destructor Documentation

__init__()

infrared.infrared.MultiDimensionalBoltzmannSampler.__init__	(		self,
			model,
			td_factory = TreeDecompositionFactory(),
			lazy = True
	)

Construct with model and optional td_factory.

Parameters

model	[const] Constraint network Model
td_factory	Factory for tree decomposition
lazy	delay construction of the data structures until required. Defaults to True

       BoltzmannSampler.__init__()

Reimplemented from infrared.infrared.BoltzmannSampler.

Member Function Documentation

is_good_sample()

infrared.infrared.MultiDimensionalBoltzmannSampler.is_good_sample	(		self,
			features,
			values
	)

whether the sample is of good quality

   checks whether the sample approximately meets the targets;
   check only the targeted features (which have value, target and tolerance)

Parameters

features

dictionary of features

rmsd()

infrared.infrared.MultiDimensionalBoltzmannSampler.rmsd (   means,   features  )

static

set_target()

infrared.infrared.MultiDimensionalBoltzmannSampler.set_target	(		self,
			target,
			tolerance,
			featureid
	)

Set target of a feature.

Parameters

target	the target value
tolerance	the tolerance (as absolute difference) to the target
fetureid	id of the feature

targeted_sample()

infrared.infrared.MultiDimensionalBoltzmannSampler.targeted_sample

(

self

)

targeted_samples()

infrared.infrared.MultiDimensionalBoltzmannSampler.targeted_samples

(

self

)

Generator of targeted samples.

   Performs multi-dimensional Boltzmann sampling: every
   self.samples_per_round many samples, the feature means are
   estimated and the weights are recalibrated. Each generated
   sample is tested for falling into target +/- tolerance for all
   features, in which case it is yielded.

   self.tweak_factor controls the scale of weight recalibration in each round.

   self.cooling is a multiplier for the tweak factor to reduce it when
   the rmsd of means to target increases (with the purpose to stabilize
   the optimization)

   self.callback callback function, called after each round with
   round statistics (FeatureStatistics)

Member Data Documentation

_targeted_samples

infrared.infrared.MultiDimensionalBoltzmannSampler._targeted_samples

protected

callback

infrared.infrared.MultiDimensionalBoltzmannSampler.callback

cooling

infrared.infrared.MultiDimensionalBoltzmannSampler.cooling

samples_per_round

infrared.infrared.MultiDimensionalBoltzmannSampler.samples_per_round

tweak_factor

infrared.infrared.MultiDimensionalBoltzmannSampler.tweak_factor

verbose

infrared.infrared.MultiDimensionalBoltzmannSampler.verbose

---

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

• src/infrared/infrared.py