-
Notifications
You must be signed in to change notification settings - Fork 1
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Merge pull request #1 from invenia/bc/prototype
Implement KeyedDistribution and KeyedSampleable
- Loading branch information
Showing
6 changed files
with
437 additions
and
9 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,5 +1,179 @@ | ||
| module KeyedDistributions | ||
|
|
||
| # Write your package code here. | ||
| using AutoHashEquals | ||
| using AxisKeys | ||
| using Distributions | ||
| using IterTools | ||
| using Random: AbstractRNG | ||
|
|
||
| export KeyedDistribution, KeyedSampleable | ||
| export axiskeys, distribution | ||
|
|
||
|
|
||
| # Constructors | ||
|
|
||
| for T in (:Distribution, :Sampleable) | ||
| KeyedT = Symbol(:Keyed, T) | ||
| @eval begin | ||
| """ | ||
| $($KeyedT)(d<:$($T), keys::Tuple{Vararg{AbstractVector}}) | ||
| Stores `keys` for each variate alongside the `$($T)` `d`, | ||
| supporting all of the common functions of a `$($T)`. | ||
| Common functions that return an `AbstractArray`, such as `rand`, | ||
| will return a `KeyedArray` with keys derived from the `$($T)`. | ||
| The type of `keys` is restricted to be consistent with | ||
| [AxisKeys.jl](https://github.com/mcabbott/AxisKeys.jl). | ||
| The length of the `keys` tuple must be the number of dimensions, which is 1 for | ||
| univariate and multivariate distributions, and 2 for matrix-variate distributions. | ||
| The length of each key vector in must match the length along each dimension. | ||
| """ | ||
| @auto_hash_equals struct $KeyedT{F<:VariateForm, S<:ValueSupport, D<:$T{F, S}} <: $T{F, S} | ||
| d::D | ||
| keys::Tuple{Vararg{AbstractVector}} | ||
|
|
||
| function $KeyedT(d::$T{F, S}, keys::Tuple{Vararg{AbstractVector}}) where {F, S} | ||
| key_lengths = map(length, keys) | ||
| key_lengths == _size(d) || throw(ArgumentError( | ||
| "lengths of key vectors $key_lengths must match " * | ||
| "size of distribution $(_size(d))")) | ||
|
|
||
| return new{F, S, typeof(d)}(d, keys) | ||
| end | ||
| end | ||
|
|
||
| """ | ||
| $($KeyedT)(d<:$($T), keys::AbstractVector) | ||
| Constructor for [`$($KeyedT)`](@ref) with one dimension of variates. | ||
| The elements of `keys` correspond to the variates of the distribution. | ||
| """ | ||
| $KeyedT(d::$T{F, S}, keys::AbstractVector) where {F, S} = $KeyedT(d, (keys, )) | ||
| end | ||
| end | ||
|
|
||
| _size(d) = (length(d),) | ||
| _size(d::Sampleable{<:Matrixvariate}) = size(d) | ||
|
|
||
| """ | ||
| KeyedDistribution(d::Distribution) | ||
| Constructs a [`KeyedDistribution`](@ref) using the keys of the first field stored in `d`, | ||
| or if there are no keys, `1:n` for the length `n` of each dimension. | ||
| """ | ||
| function KeyedDistribution(d::Distribution) | ||
| first_field = getfield(d, 1) | ||
| return KeyedDistribution(d, _keys(first_field)) | ||
| end | ||
|
|
||
| _keys(x::KeyedArray) = axiskeys(x) | ||
| _keys(x) = map(Base.OneTo, size(x)) | ||
|
|
||
| const KeyedDistOrSampleable = Union{KeyedDistribution, KeyedSampleable} | ||
|
|
||
| # Access methods | ||
|
|
||
| """ | ||
| distribution(::KeyedDistribution) -> Distribution | ||
| distribution(::KeyedSampleable{F, S, D}) -> D | ||
| Return the wrapped distribution. | ||
| """ | ||
| distribution(d::KeyedDistOrSampleable) = d.d | ||
|
|
||
| # AxisKeys functionality | ||
|
|
||
| """ | ||
| axiskeys(d::Union{KeyedDistribution, KeyedSampleable}) | ||
| Return the keys for the variates of the `KeyedDistribution` or `KeyedSampleable`. | ||
| """ | ||
| AxisKeys.axiskeys(d::KeyedDistOrSampleable) = d.keys | ||
|
|
||
| # Standard functions to overload for new Distribution and/or Sampleable | ||
| # https://juliastats.org/Distributions.jl/latest/extends/#Create-New-Samplers-and-Distributions | ||
|
|
||
| function Distributions._rand!( | ||
| rng::AbstractRNG, | ||
| d::KeyedDistOrSampleable, | ||
| x::AbstractVector{<:Real} | ||
| ) | ||
| sample = Distributions._rand!(rng, distribution(d), x) | ||
| return KeyedArray(sample, axiskeys(d)) | ||
| end | ||
|
|
||
| Base.length(d::KeyedDistOrSampleable) = length(distribution(d)) | ||
|
|
||
| Distributions.size(d::KeyedDistribution{<:Matrixvariate}) = size(distribution(d)) | ||
|
|
||
| Distributions.sampler(d::KeyedDistribution) = sampler(distribution(d)) | ||
|
|
||
| Base.eltype(d::KeyedDistribution) = eltype(distribution(d)) | ||
|
|
||
| function Distributions._logpdf(d::KeyedDistribution, x::AbstractArray) | ||
| # Workaround when KeyedArray is parameter of Distribution | ||
| # https://github.com/mcabbott/AxisKeys.jl/issues/54 | ||
| dist = distribution(d) | ||
| T = typeof(dist) | ||
| args = map(_maybe_parent, fieldvalues(dist)) | ||
| unkeyed_dist = T.name.wrapper(args...) | ||
|
|
||
| return Distributions._logpdf(unkeyed_dist, x) | ||
| end | ||
|
|
||
| _maybe_parent(x) = x | ||
| _maybe_parent(x::AbstractArray) = parent(x) | ||
|
|
||
| # Also need to overload `rand` methods to return KeyedArrays: | ||
|
|
||
| function Base.rand(rng::AbstractRNG, d::KeyedDistOrSampleable) | ||
| sample = rand(rng, distribution(d)) | ||
| ndims(sample) == 0 && return sample # univariate returns a Number | ||
| return KeyedArray(sample, axiskeys(d)) | ||
| end | ||
|
|
||
| function Base.rand(rng::AbstractRNG, d::KeyedDistOrSampleable, n::Int) | ||
| samples = rand(rng, distribution(d), n) | ||
| ndims(samples) == 1 && return KeyedArray(samples, Base.OneTo(n)) # univariate | ||
| return KeyedArray(samples, (first(axiskeys(d)), Base.OneTo(n))) | ||
| end | ||
|
|
||
| function Base.rand(rng::AbstractRNG, d::KeyedDistribution{<:Matrixvariate}, n::Int) | ||
| # Distributions.rand returns a vector of matrices | ||
| samples = [KeyedArray(x, axiskeys(d)) for x in rand(rng, distribution(d), n)] | ||
| return KeyedArray(samples, Base.OneTo(n)) | ||
| end | ||
|
|
||
| # Statistics functions for Distribution | ||
|
|
||
| Distributions.mean(d::KeyedDistribution) = KeyedArray(mean(distribution(d)), axiskeys(d)) | ||
|
|
||
| Distributions.var(d::KeyedDistribution) = KeyedArray(var(distribution(d)), axiskeys(d)) | ||
|
|
||
| function Distributions.cov(d::KeyedDistribution) | ||
| keys = vcat(axiskeys(d)...) | ||
| return KeyedArray(cov(distribution(d)), (keys, keys)) | ||
| end | ||
|
|
||
| Distributions.entropy(d::KeyedDistribution) = entropy(distribution(d)) | ||
| Distributions.entropy(d::KeyedDistribution, b::Real) = entropy(distribution(d), b) | ||
|
|
||
| # Univariate Distributions only | ||
|
|
||
| for f in (:logpdf, :quantile, :mgf, :cf) | ||
| @eval Distributions.$f(d::KeyedDistribution{<:Univariate}, x) = $f(distribution(d), x) | ||
| end | ||
|
|
||
| for f in (:minimum, :maximum, :modes, :mode, :skewness, :kurtosis) | ||
| @eval Distributions.$f(d::KeyedDistribution{<:Univariate}) = $f(distribution(d)) | ||
| end | ||
|
|
||
| # Needed to avoid method ambiguity | ||
| Distributions.cdf(d::KeyedDistribution{<:Univariate}, x::Real) = cdf(distribution(d), x) | ||
|
|
||
| function Distributions.insupport(d::KeyedDistribution{<:Univariate}, x::Real) | ||
| return insupport(distribution(d), x) | ||
| end | ||
|
|
||
| end |
Oops, something went wrong.
92bb274There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@JuliaRegistrator register
92bb274There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Registration pull request created: JuliaRegistries/General/32083
After the above pull request is merged, it is recommended that a tag is created on this repository for the registered package version.
This will be done automatically if the Julia TagBot GitHub Action is installed, or can be done manually through the github interface, or via: