de.up.ling.irtg.sampling

Class AdaptiveImportanceSampler

java.lang.Object

de.up.ling.irtg.sampling.AdaptiveImportanceSampler

public class AdaptiveImportanceSampler
extends Object

This class implements importance sampling from a tree automaton according to a user specified target distribution that does not have to be expressible by a tree automaton. The importance sampling is adaptive and attempts to optimize the proposal distribution.

Author:

teichmann

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	AdaptiveImportanceSampler.Configuration A configuration that contains all the relevant information needed to generate a number of samples from a group of tree automata.

Constructor Summary

Constructors

Constructor and Description
AdaptiveImportanceSampler() Create a new instance that will use the current time to initialize the random number generation.
AdaptiveImportanceSampler(long seed) Create a new instance that will use the given seed to initialize the random number generation.

Method Summary

Modifier and Type	Method and Description
List<TreeSample<Rule>>	adaSample(int rounds, int populationSize, RuleWeighting rw, boolean deterministic, boolean reset) Runs the sampler for the given number of rounds with the given population size and adapting the given rule weights.
TreeSample<Rule>	adaSampleMinimal(int rounds, int populationSize, RuleWeighting rw, boolean deterministic, boolean reset) Runs the sampler for the given number of rounds with the given population size and adapting the given rule weights.

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

AdaptiveImportanceSampler

public AdaptiveImportanceSampler(long seed)

Create a new instance that will use the given seed to initialize the random number generation.

Parameters:

seed -

AdaptiveImportanceSampler

public AdaptiveImportanceSampler()

Create a new instance that will use the current time to initialize the random number generation.

Method Detail

adaSample

public List<TreeSample<Rule>> adaSample(int rounds,
                                        int populationSize,
                                        RuleWeighting rw,
                                        boolean deterministic,
                                        boolean reset)

Runs the sampler for the given number of rounds with the given population size and adapting the given rule weights.

Parameters:

rounds - how many rounds of adaption we run

populationSize - the population size used during sampling.

rw - the rule weighting that gives the proposal automaton and the target weights

deterministic - whether the underlying tree automaton can be assumed to be deterministic

reset - whether we reset the rule weighting before we start sampling

Returns:

A list of tree samples - one for each adaption round - the samples can be normalized but generally have not when the result has been returned.

adaSampleMinimal

public TreeSample<Rule> adaSampleMinimal(int rounds,
                                         int populationSize,
                                         RuleWeighting rw,
                                         boolean deterministic,
                                         boolean reset)

Runs the sampler for the given number of rounds with the given population size and adapting the given rule weights.

Parameters:

rounds - how many rounds of adaption we run

populationSize - the population size used during sampling.

rw - the rule weighting that gives the proposal automaton and the target weights

deterministic - whether the underlying tree automaton can be assumed to be deterministic

reset - whether we reset the rule weighting before we start sampling

Returns:

The final tree sample the sample can be normalized but generally has not when the result has been returned.