[Serializable]
[DataContract]
[Quality(QualityBand.Experimental)]
public abstract class SequenceDistribution<TSequence, TElement, TElementDistribution, TSequenceManipulator, TAutomaton, TWeightFunction, TWeightFunctionFactory, TThis> : IDistribution<TSequence>, IDistribution, ICloneable, Diffable, SettableToUniform, HasPoint<TSequence>, CanGetLogProb<TSequence>, SettableTo<TThis>, SettableToProduct<TThis>, SettableToProduct<TThis, TThis>, SettableToRatio<TThis>, SettableToRatio<TThis, TThis>, SettableToPower<TThis>, CanGetLogAverageOf<TThis>, CanGetLogAverageOfPower<TThis>, CanGetAverageLog<TThis>, SettableToWeightedSumExact<TThis>, SettableToWeightedSum<TThis>, SettableToPartialUniform<TThis>, CanGetLogNormalizer, Sampleable<TSequence> where TSequence : class, IEnumerable<TElement> where TElementDistribution : IImmutableDistribution<TElement, TElementDistribution>, CanGetLogAverageOf<TElementDistribution>, CanComputeProduct<TElementDistribution>, CanCreatePartialUniform<TElementDistribution>, SummableExactly<TElementDistribution>, Sampleable<TElement>, new()
    where TSequenceManipulator : ISequenceManipulator<TSequence, TElement>, new()
    where TAutomaton : Automaton<TSequence, TElement, TElementDistribution, TSequenceManipulator, TAutomaton>, new()
    where TWeightFunction : WeightFunctions<TSequence, TElement, TElementDistribution, TSequenceManipulator, TAutomaton>.IWeightFunction<TWeightFunction>, new()
    where TWeightFunctionFactory : WeightFunctions<TSequence, TElement, TElementDistribution, TSequenceManipulator, TAutomaton>.IWeightFunctionFactory<TWeightFunction>, new()
    where TThis : SequenceDistribution<TSequence, TElement, TElementDistribution, TSequenceManipulator, TAutomaton, TWeightFunction, TWeightFunctionFactory, TThis>, new()

Type Parameters

Name	Description
TSequence	The type of a sequence.
TElement	The type of a sequence element.
TElementDistribution	The type of a distribution over sequence elements.
TSequenceManipulator	The type providing ways to manipulate sequences.
TAutomaton	The type of a weighted finite state automaton that can be used to represent every valid function mapping sequences to weights.
TWeightFunction	The type of an underlying function mapping sequences to weights.
TWeightFunctionFactory	The type of the factory for `TWeightFunction`.
TThis	The type of a concrete distribution class.

Constructors

SequenceDistribution()

Initializes a new instance of the SequenceDistribution<TSequence, TElement, TElementDistribution, TSequenceManipulator, TAutomaton, TWeightFunction, TWeightFunctionFactory, TThis> class with a null weight function. The weight function must be set by the subclass constructor or factory method.

Declaration

protected SequenceDistribution()

Properties

IsEmpty

Gets a value indicating whether the current distribution puts all probability mass on the empty sequence.

Declaration

public bool IsEmpty { get; }

Property Value

Type	Description
Boolean

IsPointMass

Gets a value indicating whether the current distribution represents a point mass.

Declaration

public bool IsPointMass { get; }

Property Value

Type	Description
Boolean

Point

Gets or sets the point mass represented by the distribution.

Declaration

public TSequence Point { get; set; }

Property Value

Type	Description
TSequence

UsesAutomatonRepresentation

Gets a value indicating whether the current distribution is represented as an automaton internally.

Declaration

public bool UsesAutomatonRepresentation { get; }

Property Value

Type	Description
Boolean

Methods

Any(Int32, Nullable<Int32>)

Creates a uniform distribution over sequences of length within the given bounds. If maxLength is set to null, there will be no upper bound on the length, and the resulting distribution will thus be improper.

Declaration

public static TThis Any(int minLength = 0, int? maxLength = null)

Parameters

Type	Name	Description
Int32	minLength	The minimum possible sequence length.
Nullable<Int32>	maxLength	The maximum possible sequence length, or null for no upper bound on length. Defaults to null.

Returns

Type	Description
TThis	The created distribution.

Append(TSequence, Int32)

Creates a distribution over concatenations of sequences from the current distribution and a given sequence.

Declaration

public TThis Append(TSequence sequence, int group = 0)

Parameters

Type	Name	Description
TSequence	sequence	The sequence to append.
Int32	group	The group for the appended sequence.

Returns

Type	Description
TThis	The distribution over the concatenations of sequences.

Remarks

The result is equivalent to the distribution produced by the following sampling procedure:

Sample a random sequence from the current distribution.
Append sequence to it and output the result.

Append(TElement, Int32)

Creates a distribution over concatenations of sequences from the current distribution and a given element.

Declaration

public TThis Append(TElement element, int group = 0)

Parameters

Type	Name	Description
TElement	element	The element to append.
Int32	group	The group for the appended element.

Returns

Type	Description
TThis	The distribution over the concatenations of sequences and the element.

Remarks

The result is equivalent to the distribution produced by the following sampling procedure:

Sample a random sequence from the current distribution.
Append the given element to the sampled sequence and output the result.

Append(TElementDistribution, Int32)

Creates a distribution over concatenations of sequences from the current distribution and elements from a given distribution.

Declaration

public TThis Append(TElementDistribution elementDistribution, int group = 0)

Parameters

Type	Name	Description
TElementDistribution	elementDistribution	The distribution to generate the elements from.
Int32	group	The group for the appended element.

Returns

Type	Description
TThis	The distribution over the concatenations of sequences and elements.

Remarks

The result is equivalent to the distribution produced by the following sampling procedure:

Sample a random sequence from the current distribution.
2) Sample a random element from elementDistribution.
3) Append the sampled element to the sampled sequence and output the result.

Append(TThis, Int32)

Creates a distribution over concatenations of sequences from the current distribution and sequences from a given distribution.

Declaration

public TThis Append(TThis dist, int group = 0)

Parameters

Type	Name	Description
TThis	dist	The distribution over the sequences to append.
Int32	group	The group for the appended sequence.

Returns

Type	Description
TThis	The distribution over the concatenations of sequences.

Remarks

The result is equivalent to the distribution produced by the following sampling procedure:

Sample a random sequence from the current distribution.
Sample a random sequence from dist.
Output the concatenation of the sampled pair of sequences.

AppendInPlace(TSequence, Int32)

Replaces the current distribution by a distribution over concatenations of sequences from the current distribution and a given sequence.

Declaration

public void AppendInPlace(TSequence sequence, int group = 0)

Parameters

Type	Name	Description
TSequence	sequence	The sequence to append.
Int32	group	The group for the appended sequence.

Remarks

The result is equivalent to the distribution produced by the following sampling procedure:

Sample a random sequence from the current distribution.
Append sequence to it and output the result.

AppendInPlace(TElement, Int32)

Replaces the current distribution by a distribution over concatenations of sequences from the current distribution and a given element.

Declaration

public void AppendInPlace(TElement element, int group = 0)

Parameters

Type	Name	Description
TElement	element	The element to append.
Int32	group	The group for the appended element.

Remarks

The result is equivalent to the distribution produced by the following sampling procedure:

Sample a random sequence from the current distribution.
Append the given element to the sampled sequence and output the result.

AppendInPlace(TElementDistribution, Int32)

Replaces the current distribution by a distribution over concatenations of sequences from the current distribution and elements from a given distribution.

Declaration

public void AppendInPlace(TElementDistribution elementDistribution, int group = 0)

Parameters

Type	Name	Description
TElementDistribution	elementDistribution	The distribution to generate the elements from.
Int32	group	The group for the appended element.

Remarks

The result is equivalent to the distribution produced by the following sampling procedure:

Sample a random sequence from the current distribution.
Sample a random element from elementDistribution.
Append the sampled element to the sampled sequence and output the result.

AppendInPlace(TThis, Int32)

Replaces the current distribution by a distribution over concatenations of sequences from the current distribution and sequences from a given distribution.

Declaration

public void AppendInPlace(TThis dist, int group = 0)

Parameters

Type	Name	Description
TThis	dist	The distribution over the sequences to append.
Int32	group	The group for the appended sequence.

Remarks

The result is equivalent to the distribution produced by the following sampling procedure:

Sample a random sequence from the current distribution.
Sample a random sequence from dist.
Output the concatenation of the sampled pair of sequences.

ApplyTransducer<TTransducer>(TTransducer)

Computes a distribution g(b) = sum_a f(a) T(a, b), where f(a) is the current distribution and T(a, b) is a given transducer.

Declaration

public TThis ApplyTransducer<TTransducer>(TTransducer transducer)
    where TTransducer : Transducer<TSequence, TElement, TElementDistribution, TSequenceManipulator, TAutomaton, TTransducer>, new()

Parameters

Type	Name	Description
TTransducer	transducer	The transducer to project on.

Returns

Type	Description
TThis	The projection.

Type Parameters

Name	Description
TTransducer

Clone()

Creates a copy of the current distribution.

Declaration

public TThis Clone()

Returns

Type	Description
TThis	The created copy.

Concatenate(IEnumerable<TElementDistribution>, Boolean, Boolean)

Creates a distribution over sequences induced by a given list of distributions over sequence elements where the sequence can optionally end at any length, and the last element can optionally repeat without limit.

Declaration

public static TThis Concatenate(IEnumerable<TElementDistribution> elementDistributions, bool allowEarlyEnd = false, bool repeatLastElement = false)

Parameters

Type	Name	Description
IEnumerable<TElementDistribution>	elementDistributions	Enumerable of distributions over sequence elements and the transition weights.
Boolean	allowEarlyEnd	Allow the sequence to end at any point.
Boolean	repeatLastElement	Repeat the last element.

Returns

Type	Description
TThis	The created distribution.

Contains(TSequence)

Returns true if the given sequence is in the support of this distribution (i.e. has non-zero probability).

Declaration

public bool Contains(TSequence sequence)

Parameters

Type	Name	Description
TSequence	sequence	The sequence to check.

Returns

Type	Description
Boolean	True if the sequence has non-zero probability under this distribiton, false otherwise.

Converge(TThis, Double)

Converges an improper sequence distribution

Declaration

public static TThis Converge(TThis dist, double decayWeight = 0.99)

Parameters

Type	Name	Description
TThis	dist	The original distribution.
Double	decayWeight	The decay weight.

Returns

Type	Description
TThis	The converged distribution.

Empty()

Creates a distribution which puts all probability mass on the empty sequence.

Declaration

[Construction(UseWhen = "IsEmpty")]
public static TThis Empty()

Returns

Type	Description
TThis	The created distribution.

EnumerateComponents()

Enumerates components of this distribution.

Declaration

public IEnumerable<Tuple<List<TElementDistribution>, double>> EnumerateComponents()

Returns

Type	Description
IEnumerable<Tuple<List<TElementDistribution>, Double>>	For each component, the list of element distributions and the log mixture weight for that component.

Remarks

Any sequence distribution can be represented as a mixture of 'simple' sequence distributions where each element is independent. This method enumerates through the components of this distribution treated as such a mixture. Each component is returned as a list of the independent element distributions for each element.

For some distributions, the number of components may be very large.

EnumerateSupport(Int32, Boolean)

Enumerates support of this distribution when possible. Only point mass elements are supported.

Declaration

public IEnumerable<TSequence> EnumerateSupport(int maxCount = 1000000, bool tryDeterminize = true)

Parameters

Type	Name	Description
Int32	maxCount	The maximum support enumeration count.
Boolean	tryDeterminize	Try to determinize if this distribution uses a string automaton representation.

Returns

Type	Description
IEnumerable<TSequence>	The strings supporting this distribution

Exceptions

Type	Condition
AutomatonException	Thrown if enumeration is too large.

Equals(Object)

Checks if obj equals to this distribution (i.e. represents the same distribution over sequences).

Declaration

public override bool Equals(object obj)

Parameters

Type	Name	Description
Object	obj	The object to compare this distribution with.

Returns

Type	Description
Boolean	true if this distribution is equal to `obj`, false otherwise.

Overrides

Object.Equals(Object)

FromWeightFunction(TAutomaton)

Creates a distribution from a given weight (non-normalized probability) function.

Declaration

[Construction(new string[]{"ToAutomaton"}, UseWhen = "UsesAutomatonRepresentation")]
public static TThis FromWeightFunction(TAutomaton weightFunction)

Parameters

Type	Name	Description
TAutomaton	weightFunction

Returns

Type	Description
TThis	The created distribution.

FromWeightFunction(TWeightFunction)

Creates a distribution from a given weight (non-normalized probability) function.

Declaration

[Construction(new string[]{"GetWeightFunction"})]
public static TThis FromWeightFunction(TWeightFunction weightFunction)

Parameters

Type	Name	Description
TWeightFunction	weightFunction	The weight function specifying the distribution.

Returns

Type	Description
TThis	The created distribution.

GetAverageLog(TThis)

Computes the expected logarithm of a given distribution under this distribution. Not currently supported.

Declaration

public double GetAverageLog(TThis that)

Parameters

Type	Name	Description
TThis	that	The distribution to take the logarithm of.

Returns

Type	Description
Double	`sum_x this.Evaluate(x)*Math.Log(that.Evaluate(x))`

Remarks

This is also known as the cross entropy.

GetGroups()

Declaration

public Dictionary<int, TThis> GetGroups()

Returns

Type	Description
Dictionary<Int32, TThis>

GetHashCode()

Gets the hash code of this distribution.

Declaration

public override int GetHashCode()

Returns

Type	Description
Int32	The hash code.

Overrides

Object.GetHashCode()

GetLogAverageOf(TThis)

Returns the logarithm of the probability that the current distribution would draw the same sample as a given one.

Declaration

public double GetLogAverageOf(TThis that)

Parameters

Type	Name	Description
TThis	that	The given distribution.

Returns

Type	Description
Double	The logarithm of the probability that distributions would draw the same sample.

GetLogAverageOfPower(TThis, Double)

Computes the log-integral of one distribution times another raised to a power.

Declaration

public double GetLogAverageOfPower(TThis that, double power)

Parameters

Type	Name	Description
TThis	that	The other distribution
Double	power	The exponent

Returns

Type	Description
Double	`Math.Log(sum_x this.Evaluate(x) * Math.Pow(that.Evaluate(x), power))`

Remarks

This is not the same as GetLogAverageOf(that^power) because it includes the normalization constant of that.

Powers other than 1 are not currently supported.

GetLogNormalizer()

Returns the logarithm of the normalizer of the exponential family representation of this distribution. Normalizer of an improper distribution is defined to be 1.

Declaration

public double GetLogNormalizer()

Returns

Type	Description
Double	The logarithm of the normalizer.

Remarks

Getting the normalizer is currently supported for improper distributions only.

GetLogProb(TSequence)

Gets the logarithm of the probability of a given sequence under this distribution. If the distribution is improper, returns the logarithm of the value of the underlying unnormalized weight function.

Declaration

public double GetLogProb(TSequence sequence)

Parameters

Type	Name	Description
TSequence	sequence	The sequence to get the probability for.

Returns

Type	Description
Double	The logarithm of the probability of the sequence.

GetNormalizedWeightFunction()

Returns the underlying weight function.

Declaration

public TWeightFunction GetNormalizedWeightFunction()

Returns

Type	Description
TWeightFunction	The underlying weight function.

Remarks

The returned weight function is guaranteed to be normalized. If you don't care about an arbitrary scale factor, consider using GetWeightFunction() which has less overhead.

GetProb(TSequence)

Gets the probability of a given sequence under this distribution. If the distribution is improper, returns the value of the underlying unnormalized weight function.

Declaration

public double GetProb(TSequence sequence)

Parameters

Type	Name	Description
TSequence	sequence	The sequence to get the probability for.

Returns

Type	Description
Double	The probability of the sequence.

GetWeightFunction()

Returns the underlying weight function.

Declaration

public TWeightFunction GetWeightFunction()

Returns

Type	Description
TWeightFunction	The underlying weight function.

Remarks

The returned weight function might differ from the probability function by an arbitrary scale factor. To ensure that the weight function is normalized, use GetNormalizedWeightFunction().

HasGroup(Int32)

Declaration

public bool HasGroup(int group)

Parameters

Type	Name	Description
Int32	group

Returns

Type	Description
Boolean

IsPartialUniform()

Checks whether the distribution is uniform over its support.

Declaration

public bool IsPartialUniform()

Returns

Type	Description
Boolean	true if the distribution is uniform over its support, false otherwise.

IsProper()

Checks whether the current distribution is proper.

Declaration

public bool IsProper()

Returns

Type	Description
Boolean	true if the current distribution is proper, false otherwise.

IsUniform()

Checks whether the current distribution is uniform over all possible sequences.

Declaration

public bool IsUniform()

Returns

Type	Description
Boolean	true if the current distribution is uniform over all possible sequences, false otherwise.

IsZero()

Gets a value indicating whether the current distribution is an improper distribution which assigns zero probability to every sequence.

Declaration

public bool IsZero()

Returns

Type	Description
Boolean	true if the current distribution is an improper distribution that assigns zero probability to every sequence, false otherwise.

MaxDiff(Object)

Gets a value indicating how close this distribution is to a given one in terms of probabilities they assign to sequences.

Declaration

public double MaxDiff(object that)

Parameters

Type	Name	Description
Object	that	The other distribution.

Returns

Type	Description
Double	A non-negative value, which is close to zero if the two distribution assign similar values to all sequences.

OneOf(TSequence[])

Creates a distribution which is uniform over a given set of sequences.

Declaration

public static TThis OneOf(params TSequence[] sequences)

Parameters

Type	Name	Description
TSequence[]	sequences	The set of sequences to create a distribution from.

Returns

Type	Description
TThis	The created distribution.

OneOf(TThis[])

Creates a distribution which is a uniform mixture of a given set of distributions.

Declaration

public static TThis OneOf(params TThis[] distributions)

Parameters

Type	Name	Description
TThis[]	distributions	The set of distributions to create a mixture from.

Returns

Type	Description
TThis	The created mixture distribution.

OneOf(IEnumerable<TSequence>)

Creates a distribution which is uniform over a given set of sequences.

Declaration

public static TThis OneOf(IEnumerable<TSequence> sequences)

Parameters

Type	Name	Description
IEnumerable<TSequence>	sequences	The set of sequences to create a distribution from.

Returns

Type	Description
TThis	The created distribution.

OneOf(IEnumerable<TThis>, Boolean, Boolean)

Creates a distribution which is a uniform mixture of a given set of distributions.

Declaration

public static TThis OneOf(IEnumerable<TThis> distributions, bool cloneIfSizeOne = false, bool skipNormalization = false)

Parameters

Type	Name	Description
IEnumerable<TThis>	distributions	The set of distributions to create a mixture from.
Boolean	cloneIfSizeOne	If the set has one element, whether to return that element or to clone it.
Boolean	skipNormalization	Allow normalization to be skipped as it can be expensive

Returns

Type	Description
TThis	The created mixture distribution.

OneOf(IEnumerable<KeyValuePair<TSequence, Double>>)

Creates a distribution which assigns specified probabilities to given sequences. Probabilities do not have to be normalized.

Declaration

public static TThis OneOf(IEnumerable<KeyValuePair<TSequence, double>> sequenceProbPairs)

Parameters

Type	Name	Description
IEnumerable<KeyValuePair<TSequence, Double>>	sequenceProbPairs	A list of (sequence, probability) pairs.

Returns

Type	Description
TThis	The created distribution.

OneOf(Double, TThis, Double, TThis)

Creates a mixture of a given pair of distributions.

Declaration

public static TThis OneOf(double weight1, TThis dist1, double weight2, TThis dist2)

Parameters

Type	Name	Description
Double	weight1	The weight of the first distribution.
TThis	dist1	The first distribution.
Double	weight2	The weight of the second distribution.
TThis	dist2	The second distribution.

Returns

Type	Description
TThis	The created mixture distribution.

OneOrMore(TSequence, Nullable<Int32>)

An alias for Repeat(TThis, Int32, Nullable<Int32>) with the minimum number of repetitions set to 1.

Declaration

public static TThis OneOrMore(TSequence sequence, int? maxTimes = null)

Parameters

Type	Name	Description
TSequence	sequence	The sequence.
Nullable<Int32>	maxTimes	The maximum number of repetitions, or null for no upper bound. Defaults to null.

Returns

Type	Description
TThis	The created distribution.

OneOrMore(TElement, Nullable<Int32>)

An alias for Repeat(TElement, Int32, Nullable<Int32>, DistributionKind) with the minimum number of repetitions set to 1 and uniformity kind set to UniformOverValue.

Declaration

public static TThis OneOrMore(TElement element, int? maxTimes = null)

Parameters

Type	Name	Description
TElement	element	The element.
Nullable<Int32>	maxTimes	The maximum number of repetitions, or null for no upper bound. Defaults to null.

Returns

Type	Description
TThis	The created distribution.

OneOrMore(TElementDistribution, Nullable<Int32>)

An alias for Repeat(TElementDistribution, Int32, Nullable<Int32>, DistributionKind) with the minimum number of repetitions set to 0 and uniformity kind set to UniformOverValue.

Declaration

public static TThis OneOrMore(TElementDistribution allowedElements, int? maxTimes = null)

Parameters

Type	Name	Description
TElementDistribution	allowedElements	The allowed sequence elements.
Nullable<Int32>	maxTimes	The maximum number of repetitions, or null for no upper bound. Defaults to null.

Returns

Type	Description
TThis	The created distribution.

OneOrMore(TThis, Nullable<Int32>)

An alias for Repeat(TThis, Int32, Nullable<Int32>) with the minimum number of repetitions set to 1.

Declaration

public static TThis OneOrMore(TThis dist, int? maxTimes = null)

Parameters

Type	Name	Description
TThis	dist	The distribution.
Nullable<Int32>	maxTimes	The maximum number of repetitions, or null for no upper bound. Defaults to null.

Returns

Type	Description
TThis	The created distribution.

Optional(TSequence, Double)

Creates a mixture of a point mass on the given sequence and a point mass representing an empty sequence.

Declaration

public static TThis Optional(TSequence sequence, double prob = 0.5)

Parameters

Type	Name	Description
TSequence	sequence	The given sequence.
Double	prob	The probability of the component corresponding to `sequence`.

Returns

Type	Description
TThis	The created mixture.

Optional(TThis, Double)

Creates a mixture of a given distribution and a point mass representing an empty sequence.

Declaration

public static TThis Optional(TThis dist, double prob = 0.5)

Parameters

Type	Name	Description
TThis	dist	The distribution.
Double	prob	The probability of the component corresponding to `dist`.

Returns

Type	Description
TThis	The created mixture.

PointMass(TSequence)

Creates a point mass distribution.

Declaration

[Construction(new string[]{"Point"}, UseWhen = "IsPointMass")]
public static TThis PointMass(TSequence point)

Parameters

Type	Name	Description
TSequence	point	The point.

Returns

Type	Description
TThis	The created point mass distribution.

Product(TThis)

Returns a product of the current distribution and a given one.

Declaration

public TThis Product(TThis that)

Parameters

Type	Name	Description
TThis	that	The distribution to compute the product with.

Returns

Type	Description
TThis	The product.

Repeat(TSequence, Int32, Nullable<Int32>)

Creates a uniform distribution over sequences containing N repeats of the specified sequence, where n lies within the given bounds. If maxTimes is set to null, there will be no upper bound on the number of repeats, and the resulting distribution will thus be improper.

Declaration

public static TThis Repeat(TSequence sequence, int minTimes = 1, int? maxTimes = null)

Parameters

Type	Name	Description
TSequence	sequence	The sequence to repeat.
Int32	minTimes	The minimum possible sequence length. Defaults to 1.
Nullable<Int32>	maxTimes	The maximum possible sequence length, or null for no upper bound on length. Defaults to null.

Returns

Type	Description
TThis	The created distribution.

Repeat(TElement, Int32, Nullable<Int32>, DistributionKind)

Creates a uniform distribution over sequences of length within the given bounds. Sequence elements are restricted to be equal to a given element. If maxTimes is set to null, there will be no upper bound on the length, and the resulting distribution will thus be improper.

Declaration

public static TThis Repeat(TElement element, int minTimes = 1, int? maxTimes = null, DistributionKind uniformity = DistributionKind.UniformOverValue)

Parameters

Type	Name	Description
TElement	element	The element.
Int32	minTimes	The minimum possible sequence length. Defaults to 1.
Nullable<Int32>	maxTimes	The maximum possible sequence length, or null for no upper bound on length. Defaults to null.
DistributionKind	uniformity	The type of uniformity.

Returns

Type	Description
TThis	The created distribution.

Repeat(TElementDistribution, Int32, Nullable<Int32>, DistributionKind)

Creates a uniform distribution over sequences of length within the given bounds. Sequence elements are restricted to be non-zero probability elements from a given distribution. If maxTimes is set to null, there will be no upper bound on the length, and the resulting distribution will thus be improper.

Declaration

public static TThis Repeat(TElementDistribution allowedElements, int minTimes = 1, int? maxTimes = null, DistributionKind uniformity = DistributionKind.UniformOverValue)

Parameters

Type	Name	Description
TElementDistribution	allowedElements	The distribution representing allowed sequence elements.
Int32	minTimes	The minimum possible sequence length. Defaults to 1.
Nullable<Int32>	maxTimes	The maximum possible sequence length, or null for no upper bound on length. Defaults to null.
DistributionKind	uniformity	The type of uniformity

Returns

Type	Description
TThis	The created distribution.

Repeat(TThis, Int32, Nullable<Int32>)

Creates a distribution by applying Repeat(TThis, Int32, Nullable<Int32>) to the weight function of a given distribution, which is additionally scaled by the inverse of GetLogAverageOf(TThis) with itself. So, if the given distribution is partial uniform, the result will be partial uniform over the repetitions of sequences covered by the distribution.

If maxTimes is set to null, there will be no upper bound on the length, and the resulting distribution will thus be improper.

Declaration

public static TThis Repeat(TThis dist, int minTimes = 1, int? maxTimes = null)

Parameters

Type	Name	Description
TThis	dist	The distribution.
Int32	minTimes	The minimum number of repetitions. Defaults to 1.
Nullable<Int32>	maxTimes	The maximum number of repetitions, or null for no upper bound. Defaults to null.

Returns

Type	Description
TThis	The created distribution.

Sample()

Draws a sample from the distribution.

Declaration

public TSequence Sample()

Returns

Type	Description
TSequence	The drawn sample.

Sample(TSequence)

Draws a sample from the distribution.

Declaration

public TSequence Sample(TSequence result)

Parameters

Type	Name	Description
TSequence	result	A pre-allocated storage for the sample (will be ignored).

Returns

Type	Description
TSequence	The drawn sample.

SetLogValueOverride(Nullable<Double>)

Sets a value that, if not null, will be returned when computing the log probability of any sequence which is in the support of this distribution.

Declaration

public void SetLogValueOverride(double? logValueOverride)

Parameters

Type	Name	Description
Nullable<Double>	logValueOverride	A non-null value that should be returned when computing the log probability of any sequence which is in the support of this distribution, or null to clear any existing override.

SetTo(TThis)

Replaces the current distribution by a copy of the given distribution.

Declaration

public void SetTo(TThis that)

Parameters

Type	Name	Description
TThis	that	The distribution to set the current distribution to.

SetToPartialUniform()

Sets the distribution to be uniform over its support.

Declaration

public void SetToPartialUniform()

Remarks

This function will fail if the weight function of this distribution is represented by an automaton that cannot be determinized.

SetToPartialUniformOf(TThis)

Sets the distribution to be uniform over the support of a given distribution.

Declaration

public void SetToPartialUniformOf(TThis dist)

Parameters

Type	Name	Description
TThis	dist	The distribution which support will be used to setup the current distribution.

Remarks

This function will fail if the weight function of dist is represented by an automaton that cannot be determinized.

SetToPower(TThis, Double)

Replaces the current distribution with a given distribution raised to a given power.

Declaration

public void SetToPower(TThis that, double power)

Parameters

Type	Name	Description
TThis	that	The distribution to raise to the power.
Double	power	The power.

Remarks

Only 0 and 1 are currently supported as powers.

SetToProduct(TThis, TThis)

Replaces the current distribution with a product of a given pair of distributions.

Declaration

public void SetToProduct(TThis dist1, TThis dist2)

Parameters

Type	Name	Description
TThis	dist1	The first distribution.
TThis	dist2	The second distribution.

SetToProductAndReturnLogNormalizer(TThis, TThis, Boolean)

Replaces the current distribution with a product of a given pair of distributions Returns the logarithm of the normalizer for the product (as returned by GetLogAverageOf(TThis)).

Declaration

public double SetToProductAndReturnLogNormalizer(TThis dist1, TThis dist2, bool tryDeterminize = true)

Parameters

Type	Name	Description
TThis	dist1	The first distribution.
TThis	dist2	The second distribution.
Boolean	tryDeterminize	Whether to try to determinize the result.

Returns

Type	Description
Double	The logarithm of the normalizer for the product.

SetToRatio(TThis, TThis, Boolean)

Replaces the current distribution with the ratio of a given pair of distributions. Not currently supported.

Declaration

public void SetToRatio(TThis numerator, TThis denominator, bool forceProper)

Parameters

Type	Name	Description
TThis	numerator	The numerator in the ratio.
TThis	denominator	The denominator in the ratio.
Boolean	forceProper	Specifies whether the ratio must be proper.

SetToSum(Double, TThis, Double, TThis)

Replaces the current distribution with a mixture of a given pair of distributions.

Declaration

public void SetToSum(double weight1, TThis dist1, double weight2, TThis dist2)

Parameters

Type	Name	Description
Double	weight1	The weight of the first distribution.
TThis	dist1	The first distribution.
Double	weight2	The weight of the second distribution.
TThis	dist2	The second distribution.

SetToSumLog(Double, TThis, Double, TThis)

Replaces the current distribution with a mixture of a given pair of distributions.

Declaration

public void SetToSumLog(double logWeight1, TThis dist1, double logWeight2, TThis dist2)

Parameters

Type	Name	Description
Double	logWeight1	The logarithm of the weight of the first distribution.
TThis	dist1	The first distribution.
Double	logWeight2	The logarithm of the weight of the second distribution.
TThis	dist2	The second distribution.

SetToUniform()

Replaces the current distribution with an improper distribution which assigns the probability of 1 to every sequence.

Declaration

public void SetToUniform()

SetToUniformOf(TElementDistribution)

Replaces the current distribution with an improper distribution which assigns the probability of 1 to every sequence. Sequence elements are restricted to be non-zero probability elements from a given distribution.

Declaration

public void SetToUniformOf(TElementDistribution allowedElements)

Parameters

Type	Name	Description
TElementDistribution	allowedElements	The distribution representing allowed sequence elements.

SetToUniformOf(TElementDistribution, Double)

Replaces the current distribution with an improper distribution which assigns a given probability to every sequence. Sequence elements are restricted to be non-zero probability elements from a given distribution.

Declaration

protected void SetToUniformOf(TElementDistribution allowedElements, double uniformLogProb)

Parameters

Type	Name	Description
TElementDistribution	allowedElements	The distribution representing allowed sequence elements.
Double	uniformLogProb	The logarithm of the probability assigned to every allowed sequence.

SetToZero()

Replaces the current distribution by an improper zero distribution.

Declaration

public void SetToZero()

SetWeightFunction(TAutomaton, Boolean)

Replaces the weight function of the current distribution with a given one.

Declaration

public void SetWeightFunction(TAutomaton weightFunction, bool normalizeStructure = true)

Parameters

Type	Name	Description
TAutomaton	weightFunction
Boolean	normalizeStructure	Whether to normalize the structure of the distribution

SetWeightFunction(TWeightFunction, Boolean)

Replaces the weight function of the current distribution with a given one.

Declaration

public void SetWeightFunction(TWeightFunction weightFunction, bool normalizeStructure = true)

Parameters

Type	Name	Description
TWeightFunction	weightFunction	The weight function to replace the current one with.
Boolean	normalizeStructure	Whether to normalize the structure of the distribution

SingleElement(TElement)

Creates a distribution which puts all probability mass on a sequence containing only a given element.

Declaration

public static TThis SingleElement(TElement element)

Parameters

Type	Name	Description
TElement	element	The element.

Returns

Type	Description
TThis	The created distribution.

SingleElement(TElementDistribution)

Creates a distribution over sequences of length 1 induced by a given distribution over sequence elements.

Declaration

public static TThis SingleElement(TElementDistribution elementDistribution)

Parameters

Type	Name	Description
TElementDistribution	elementDistribution	The distribution over sequence elements.

Returns

Type	Description
TThis	The created distribution.

Remarks

The distribution created by this method can differ from the result of Repeat(TElementDistribution, Int32, Nullable<Int32>, DistributionKind) with both min and max number of times to repeat set to 1 since the latter always creates a partial uniform distribution.

ToAutomaton()

Returns an automaton representing the underlying weight function.

Declaration

public TAutomaton ToAutomaton()

Returns

Type	Description
TAutomaton	An automaton representing the underlying weight function.

Remarks

The returned weight function might differ from the probability function by an arbitrary scale factor. To ensure that the weight function is normalized, use ToNormalizedAutomaton().

ToNormalizedAutomaton()

Returns an automaton representing the underlying weight function.

Declaration

public TAutomaton ToNormalizedAutomaton()

Returns

Type	Description
TAutomaton	An automaton representing the underlying weight function.

Remarks

The returned weight function is guaranteed to be normalized. If you don't care about an arbitrary scale factor, consider using ToAutomaton() which has less overhead.

ToString()

Returns a string that represents the distribution.

Declaration

public override string ToString()

Returns

Type	Description
String	A string that represents the distribution.

Overrides

Object.ToString()

ToString(ISequenceDistributionFormat)

Returns a string that represents the distribution.

Declaration

public string ToString(ISequenceDistributionFormat format)

Parameters

Type	Name	Description
ISequenceDistributionFormat	format	The format.

Returns

Type	Description
String	A string that represents the distribution.

ToString(Action<TElementDistribution, StringBuilder>)

Returns a string that represents the distribution.

Declaration

protected string ToString(Action<TElementDistribution, StringBuilder> appendRegex)

Parameters

Type	Name	Description
Action<TElementDistribution, StringBuilder>	appendRegex	Optional method for appending at the element distribution level.

Returns

Type	Description
String	A string that represents the automaton.

TryDeterminize()

Tries to replace the internal representation of the current distribution with an equivalent deterministic automaton, unless the current representation is guaranteed to produce a deterministic automaton on conversion.

Declaration

public bool TryDeterminize()

Returns

Type	Description
Boolean	true if the internal representation of the current distribution was successfully replaced with a deterministic automaton or is guaranteed to produce one on direct conversion, false otherwise.

TryEnumerateSupport(Int32, out IEnumerable<TSequence>, Boolean)

Enumerates support of this distribution when possible. Only point mass elements are supported.

Declaration

public bool TryEnumerateSupport(int maxCount, out IEnumerable<TSequence> result, bool tryDeterminize = true)

Parameters

Type	Name	Description
Int32	maxCount	The maximum support enumeration count.
IEnumerable<TSequence>	result	The strings supporting this distribution.
Boolean	tryDeterminize	Try to determinize if this distribution uses a string automaton representation.

Returns

Type	Description
Boolean	True if successful, false otherwise.

Exceptions

Type	Condition
AutomatonException	Thrown if enumeration is too large.

Uniform()

Creates an improper distribution which assigns the probability of 1 to every sequence.

Declaration

[Construction(UseWhen = "IsUniform")]
public static TThis Uniform()

Returns

Type	Description
TThis	The created uniform distribution.

UniformOf(TElementDistribution, Double)

Creates an improper distribution which assigns a given probability to every sequence. Sequence elements are restricted to be non-zero probability elements from a given distribution.

Declaration

protected static TThis UniformOf(TElementDistribution allowedElements, double uniformLogProb)

Parameters

Type	Name	Description
TElementDistribution	allowedElements	The distribution representing allowed sequence elements.
Double	uniformLogProb	The logarithm of the probability assigned to every allowed sequence.

Returns

Type	Description
TThis	The created distribution.

Zero()

Creates an improper distribution which assigns zero probability to every sequence.

Declaration

[Construction(UseWhen = "IsZero")]
public static TThis Zero()

Returns

Type	Description
TThis	The created zero distribution.

ZeroOrMore(TSequence, Nullable<Int32>)

An alias for Repeat(TThis, Int32, Nullable<Int32>) with the minimum number of repetitions set to 0.

Declaration

public static TThis ZeroOrMore(TSequence sequence, int? maxTimes = null)

Parameters

Type	Name	Description
TSequence	sequence	The sequence.
Nullable<Int32>	maxTimes	The maximum number of repetitions, or null for no upper bound. Defaults to null.

Returns

Type	Description
TThis	The created distribution.

ZeroOrMore(TElement, Nullable<Int32>)

An alias for Repeat(TElement, Int32, Nullable<Int32>, DistributionKind) with the minimum number of repetitions set to 0 and uniformity kind set to UniformOverValue.

Declaration

public static TThis ZeroOrMore(TElement element, int? maxTimes = null)

Parameters

Type	Name	Description
TElement	element	The element.
Nullable<Int32>	maxTimes	The maximum number of repetitions, or null for no upper bound. Defaults to null.

Returns

Type	Description
TThis	The created distribution.

ZeroOrMore(TElementDistribution, Nullable<Int32>)

An alias for Repeat(TElementDistribution, Int32, Nullable<Int32>, DistributionKind) with the minimum number of repetitions set to 0 and uniformity kind set to UniformOverValue.

Declaration

public static TThis ZeroOrMore(TElementDistribution allowedElements, int? maxTimes = null)

Parameters

Type	Name	Description
TElementDistribution	allowedElements	The allowed sequence elements.
Nullable<Int32>	maxTimes	The maximum number of repetitions, or null for no upper bound. Defaults to null.

Returns

Type	Description
TThis	The created distribution.

ZeroOrMore(TThis, Nullable<Int32>)

An alias for Repeat(TThis, Int32, Nullable<Int32>) with the minimum number of repetitions set to 0.

Declaration

public static TThis ZeroOrMore(TThis dist, int? maxTimes = null)

Parameters

Type	Name	Description
TThis	dist	The distribution.
Nullable<Int32>	maxTimes	The maximum number of repetitions, or null for no upper bound. Defaults to null.

Returns

Type	Description
TThis	The created distribution.

Explicit Interface Implementations

ICloneable.Clone()

Creates a copy of the current distribution.

Declaration

object ICloneable.Clone()

Returns

Type	Description
Object	The created copy.

Implements

IDistribution<T>

IDistribution

System.ICloneable

Diffable

SettableToUniform

HasPoint<T>

CanGetLogProb<T>

SettableTo<T>

SettableToProduct<T>

SettableToProduct<T, U>

SettableToRatio<T>

SettableToRatio<T, U>

SettableToPower<T>

CanGetLogAverageOf<T>

CanGetLogAverageOfPower<T>

CanGetAverageLog<T>

SettableToWeightedSumExact<T>

SettableToWeightedSum<T>

SettableToPartialUniform<TDist>

CanGetLogNormalizer

Sampleable<T>