io.improbable.keanu.vertices.dbl.probabilistic

Class LaplaceVertex

java.lang.Object

io.improbable.keanu.vertices.Vertex<DoubleTensor>

io.improbable.keanu.vertices.dbl.DoubleVertex

io.improbable.keanu.vertices.dbl.probabilistic.LaplaceVertex

All Implemented Interfaces:

io.improbable.keanu.kotlin.DoubleOperators<DoubleVertex>, io.improbable.keanu.kotlin.NumberOperators<DoubleVertex>, Differentiable, ProbabilisticDouble, Observable<DoubleTensor>, Probabilistic<DoubleTensor>

public class LaplaceVertex
extends DoubleVertex
implements ProbabilisticDouble

Constructor Summary

Constructors

Constructor and Description
LaplaceVertex(double mu, double beta)
LaplaceVertex(double mu, DoubleVertex beta)
LaplaceVertex(DoubleVertex mu, double beta)
LaplaceVertex(DoubleVertex mu, DoubleVertex beta) One to one constructor for mapping some shape of mu and sigma to a matching shaped Laplace.
LaplaceVertex(long[] shape, double mu, double beta)
LaplaceVertex(long[] shape, double mu, DoubleVertex beta)
LaplaceVertex(long[] shape, DoubleVertex mu, double beta)
LaplaceVertex(long[] tensorShape, DoubleVertex mu, DoubleVertex beta) One mu or beta or both that match a proposed tensor shape of Laplace

Method Summary

Modifier and Type	Method and Description
java.util.Map<Vertex,DoubleTensor>	dLogProb(DoubleTensor value, java.util.Set<? extends Vertex> withRespectTo) The partial derivatives of the natural log prob.
double	logProb(DoubleTensor value) This is the natural log of the probability at the supplied value.
DoubleTensor	sample(KeanuRandom random)

Methods inherited from class io.improbable.keanu.vertices.dbl.DoubleVertex

abs, acos, asin, atan, atan2, ceil, concat, cos, div, div, divideBy, divideBy, equalTo, exp, floor, forwardModeAutoDifferentiation, getValue, greaterThan, greaterThanOrEqualTo, lambda, lambda, lessThan, lessThanOrEqualTo, log, logGamma, matrixDeterminant, matrixInverse, matrixMultiply, max, min, minus, minus, multiply, multiply, notEqualTo, observe, observe, plus, plus, pow, pow, reshape, reverseDiv, reverseMinus, reverseModeAutoDifferentiation, round, setAndCascade, setAndCascade, setValue, setValue, sigmoid, sin, slice, sum, sum, take, tan, times, times, unaryMinus

Methods inherited from class io.improbable.keanu.vertices.Vertex

addChild, addParent, addParents, equals, eval, getChildren, getConnectedGraph, getId, getIndentation, getLabel, getObservedValue, getParents, getRawValue, getShape, getValue, hashCode, hasValue, isObserved, isProbabilistic, lazyEval, observe, observeOwnValue, removeLabel, sample, setAndCascade, setLabel, setLabel, setParents, setParents, setValue, toString, unobserve

Methods inherited from class java.lang.Object

clone, finalize, getClass, notify, notifyAll, wait, wait, wait

Methods inherited from interface io.improbable.keanu.vertices.dbl.probabilistic.ProbabilisticDouble

dLogPdf, dLogPdf, dLogPdf, dLogPdf, dLogPdf, dLogPdf, logPdf, logPdf, logPdf

Methods inherited from interface io.improbable.keanu.vertices.Probabilistic

dLogProb, dLogProbAtValue, dLogProbAtValue, getValue, keepOnlyProbabilisticVertices, logProbAtValue, setValue

Methods inherited from interface io.improbable.keanu.vertices.Observable

getObservedValue, isObserved, observableTypeFor, observe, unobserve

Methods inherited from interface io.improbable.keanu.vertices.dbl.Differentiable

getDerivativeWrtLatents

Constructor Detail

LaplaceVertex

public LaplaceVertex(long[] tensorShape,
                     DoubleVertex mu,
                     DoubleVertex beta)

One mu or beta or both that match a proposed tensor shape of Laplace

If all provided parameters are scalar then the proposed shape determines the shape

Parameters:

tensorShape - the desired shape of the tensor within the vertex

mu - the mu of the Laplace with either the same shape as specified for this vertex or a scalar

beta - the beta of the Laplace with either the same shape as specified for this vertex or a scalar

LaplaceVertex

public LaplaceVertex(long[] shape,
                     DoubleVertex mu,
                     double beta)

LaplaceVertex

public LaplaceVertex(long[] shape,
                     double mu,
                     DoubleVertex beta)

LaplaceVertex

public LaplaceVertex(long[] shape,
                     double mu,
                     double beta)

LaplaceVertex

public LaplaceVertex(DoubleVertex mu,
                     DoubleVertex beta)

One to one constructor for mapping some shape of mu and sigma to a matching shaped Laplace.

Parameters:

mu - the mu of the Laplace with either the same shape as specified for this vertex or a scalar

beta - the beta of the Laplace with either the same shape as specified for this vertex or a scalar

LaplaceVertex

public LaplaceVertex(DoubleVertex mu,
                     double beta)

LaplaceVertex

public LaplaceVertex(double mu,
                     DoubleVertex beta)

LaplaceVertex

public LaplaceVertex(double mu,
                     double beta)

Method Detail

logProb

public double logProb(DoubleTensor value)

Description copied from interface: Probabilistic

This is the natural log of the probability at the supplied value. In the case of continuous vertices, this is actually the log of the density, which will differ from the probability by a constant.

Specified by:

logProb in interface Probabilistic<DoubleTensor>

Parameters:

value - The supplied value.

Returns:

The natural log of the probability function at the supplied value. For continuous variables this is called the PDF (probability density function). For discrete variables this is called the PMF (probability mass function).

dLogProb

public java.util.Map<Vertex,DoubleTensor> dLogProb(DoubleTensor value,
                                                   java.util.Set<? extends Vertex> withRespectTo)

Description copied from interface: Probabilistic

The partial derivatives of the natural log prob.

Specified by:

dLogProb in interface Probabilistic<DoubleTensor>

Parameters:

value - at a given value

withRespectTo - list of parents to differentiate with respect to

Returns:

the partial derivatives of the log of the probability function at the supplied value. For continuous variables this is called the PDF (probability density function). For discrete variables this is called the PMF (probability mass function).

sample

public DoubleTensor sample(KeanuRandom random)

Specified by:

sample in class Vertex<DoubleTensor>

Parameters:

random - source of randomness

Returns:

a sample from the vertex's distribution. For non-probabilistic vertices, this will always be the same value.