Abstract Classes

DMixLearn captures most distributions in the exponential family. A detailed walkthrough on defining a custom distribution class can be found in User Defined Classes. We list the abstract classes that exist in DMixLearn below.

ProbabilityDistribution

class dml.stats.pdist.ProbabilityDistribution

Defines ProbabilityDistribution Abstract Class.

Note

This is generally used as an inherited class for SequenceEncodableProbabilityDistribution.

__init__()

abstract estimator(pseudo_count=None)

Create a ParameterEstimator for corresponding SequenceEncodableProbabilityDistribution.

Parameters:: pseudo_count (Optional[float]) – Regularize sufficient statistics in estimation step.
Return type:: ParameterEstimator
Returns:: ParameterEstimator

abstract log_density(x)

Evaluate the log-density of distribution.

Return type:: float
Returns:: float

abstract sampler(seed=None)

Create a DistributionSampler object for a given ProbabilityDistribution.

Parameters:: seed (Optional[int]) – Set seed for drawing samples from distribution.
Return type:: DistributionSampler

SequenceEncodableProbabilityDistribution

class dml.stats.pdist.SequenceEncodableProbabilityDistribution

Extends the ProbabilityDistribution to handle vectorized calls.

abstract dist_to_encoder()

Create DataSequenceEncoder object for SequenceEncodableProbabilityDistribution instance.

Return type:: DataSequenceEncoder
Returns:: DataSequenceEncoder

abstract seq_log_density(x)

Vectorized evaluation of the log density.

Parameters:: x (EncodedDataSequence) – EncodedDataSequence for corresponding SequenceEncodedProbabilityDistribution.
Return type:: ndarray
Returns:: np.ndarray

DistributionSampler

class dml.stats.pdist.DistributionSampler(dist, seed=None)

DistributionSampler is an Abstract class for distribution samplers.

dist

Distribution to sample from.

Type:: SequenceEncodableProbabilityDistribution

rng

Random number generator.

Type:: RandomState

__init__(dist, seed=None)

Initialize DistributionSampler.

Parameters:

dist (SequenceEncodableProbabilityDistribution) – Distribution to sample from.
seed (Optional[int]) – Used to set seed on rng.

new_seed()

Generates a new seed from rng

Return type:: int

abstract sample(size=None)

Generate samples from distribution.

Parameters:: size (Optional[int]) – Number of samples to generate.
Return type:: Any
Returns:: Samples from distribution.

ConditionalSampler

class dml.stats.pdist.ConditionalSampler

AbstractClass for ConditionalSampler.

Note

This is only implemented for samples of conditional distributions.

abstract sample_given(x)

Sample at conditional value.

Parameters:: x (Any) – Conditioned on x, sample from dist.
Returns:: Sample from conditional distribution.

StatisticAccumulator

class dml.stats.pdist.StatisticAccumulator

abstract combine(suff_stat)

Method for combining aggregated sufficient statistics.

Parameters:: suff_stat (SS) – Sufficient statistics.
Return type:: StatisticAccumulator
Returns:: None

abstract from_value(x)

Set sufficient statistics equal to passed value.

Parameters:: x (SS) – Generic sufficient statistic for instance of StatisticAccumulator.
Return type:: SequenceEncodableStatisticAccumulator

initialize(x, weight, rng)

Initialize sufficient statistics for a single data observation.

Note

Used for debugging only.

Parameters:

x (Any) – Data type corresponding to StatisticAccumulator object.
weight (float) – Weight associated with single observation.
rng (np.random.RandomState) – Set seed for initialization.

Return type:

None

abstract key_merge(stats_dict)

Merge sufficient statistics with matching keys.

Parameters:: stats_dict (Dict[str, Any]) – Dict mapping keys to sufficient statistic value or accumulator.
Return type:: None

abstract key_replace(stats_dict)

Set sufficient statistics of accumulator instance to key’d values.

Parameters:: stats_dict (Dict[str, Any]) – Dict mapping keys to sufficient statistic value or accumulator.
Return type:: None

update(x, weight, estimate)

Accumulate sufficient statistics for a single data observation.

Note

Used for debugging only.

Parameters:

x (Any) – Data type corresponding to StatisticAccumulator object.
weight (float) – Weight associated with single observation.
estimate (SequenceEncodableProbabilityDistribution) – Previous estimate of distribution.

Return type:

None

abstract value()

Return sufficient statistics of StatisticAccumulator.

Return type:: TypeVar(SS)

SequenceEncodableStatisticAccumulator

class dml.stats.pdist.SequenceEncodableStatisticAccumulator

abstract acc_to_encoder()

Create DataSequenceEncoder object for SequenceEncodableStatisticAccumulator instance.

Return type:: DataSequenceEncoder

abstract seq_initialize(x, weights, rng)

Vectorized initialization of sufficient statistics.

Parameters:

x (EncodedDataSequence) – EncodedDataSequence for given SequenceEncodableStatisticAccumulator type.
weights (np.ndarray) – weights for observations.
rng (np.random.RandomState) – RandomState object for setting seed on initialization.

Return type:

None

abstract seq_update(x, weights, estimate)

Vectorized accumulation of sufficient statistics for EM updates.

Parameters:

x (EncodedDataSequence) – EncodedDataSequence for given SequenceEncodableStatisticAccumulator type.
weights (np.ndarray) – weights for observations.
estimate (Optional[SequenceEncodableProbabilityDistribution]) – Optional previous estimate of distribution.

Return type:

None

ParameterEstimator

class dml.stats.pdist.ParameterEstimator(*args)

Abstract class for ParameterEstimator object.

abstract __init__(*args): Must implement constructor for ParameterEstimator

abstract accumulator_factory()

Create SequenceEncodableStatisticAccumulator object.

Return type:: StatisticAccumulatorFactory

abstract estimate(nobs, suff_stat)

Estimate SequenceEncodableProbabilityDistribution for sufficient statistics.

Parameters:

nobs (Optional[float]) – Weighted number of observations.
suff_stat (Tuple[int, np.ndarray, np.ndarray, np.ndarray]) – Sufficient statistics for dirichlet distribution.

Return type:

SequenceEncodableProbabilityDistribution

Returns:

SequenceEncodableProbabilityDistribution

DataSequenceEncoder

class dml.stats.pdist.DataSequenceEncoder

abstract seq_encode(x)

Create EncodedDataSequence from iid observations from SequenceEncodedProbabilityDistribution.

Parameters:: x (Any) – Sequence of observations from corresponding distribution.
Return type:: EncodedDataSequence
Returns:: EncodedDataSequence

EncodedDataSequence

class dml.stats.pdist.EncodedDataSequence(data)

EncodedDatSequence is the outputed data structure from DataSeqeunceEncoder. Object is used for vectorized functions and type checks.

__init__(data)

Create instance of EncodedDataSequence.

Parameters:: data (Any) – Store the data encocded for vectorized calls.