Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add DiscreteNonParametric distribution #634

Merged
merged 17 commits into from
Nov 4, 2018
Merged

Add DiscreteNonParametric distribution #634

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
17 commits
Select commit Hold shift + click to select a range
a7b6bb5
Add Generic discrete distribution
Jul 9, 2017
712373d
Convert modes to explicit loop
Jul 9, 2017
1aa2656
Add uniqueness check for values in Generic constructor
Jul 11, 2017
2db9851
Convert Generic's support to an AbstractVector
Jul 12, 2017
744f067
Convert Categorical to a type alias of Generic
Jul 13, 2017
0d77971
Add Generic mgf and cf
Nov 4, 2017
236d652
Add Generic fit and sufficient statistic methods
Nov 4, 2017
536f177
Easy 0.7 fixes
Sep 23, 2018
a2f0712
Updates for Random in 0.7 and tests passing
Sep 23, 2018
64dcfab
Fix 0.7 constructor dispatch
Sep 23, 2018
5688d58
Fix weighted suffstats and minor changes
Sep 24, 2018
7f803cb
Rename Generic to DiscreteNonParametric
Oct 27, 2018
9a607b9
Add documentation for DiscreteNonParametric
Oct 27, 2018
3c2efbe
Generalize probability vector type for DiscreteNonParametric and Cate…
Oct 30, 2018
677eaa6
Generalize DiscreteNonParametric/Categorical samplers
Oct 30, 2018
a11c08d
Further specialize support of Categorical
Oct 30, 2018
733434b
Partially support test_distr with DiscreteNonParametric
Oct 31, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
Generalize probability vector type for DiscreteNonParametric and Cate… 
…gorical, fix #743
  • Loading branch information
Gord Stephen committed Oct 30, 2018
commit 3c2efbe6048322769846484dcb87a86bbbb30933
38 changes: 22 additions & 16 deletions src/univariate/discrete/categorical.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,34 +24,39 @@ External links:

* [Categorical distribution on Wikipedia](http://en.wikipedia.org/wiki/Categorical_distribution)
"""
Categorical{T} = DiscreteNonParametric{Int,T,UnitRange{Int}}
Categorical{P,Ps} = DiscreteNonParametric{Int,P,UnitRange{Int},Ps}

Categorical{P}(p::Vector{P}, ::NoArgCheck) where P =
Categorical{P}(1:length(p), p, NoArgCheck())
Categorical(p::Vector{P}, ::NoArgCheck) where P = Categorical{P}(p, NoArgCheck())
Categorical{P,Ps}(p::Ps, ::NoArgCheck) where {P<:Real, Ps<:AbstractVector{P}} =
Categorical{P,Ps}(1:length(p), p, NoArgCheck())

function Categorical{P}(p::Vector{P}) where P
@check_args(DiscreteNonParametric, isprobvec(p))
Categorical{P}(1:length(p), p, NoArgCheck())
Categorical(p::Ps, ::NoArgCheck) where {P<:Real, Ps<:AbstractVector{P}} =
Categorical{P,Ps}(p, NoArgCheck())

function Categorical{P,Ps}(p::Ps) where {P<:Real, Ps<:AbstractVector{P}}
@check_args(Categorical, isprobvec(p))
Categorical{P,Ps}(1:length(p), p, NoArgCheck())
end
Categorical(p::Vector{P}) where P = Categorical{P}(p)

Categorical(p::Ps) where {P<:Real, Ps<:AbstractVector{P}} =
Categorical{P,Ps}(p)

function Categorical(k::Integer)
@check_args(DiscreteNonParametric, k >= 1)
Categorical{Float64}(1:k, fill(1/k, k), NoArgCheck())
@check_args(Categorical, k >= 1)
Categorical{Float64,Vector{Float64}}(1:k, fill(1/k, k), NoArgCheck())
end

@distr_support Categorical 1 support(d).stop

### Conversions

convert(::Type{Categorical{T}}, p::Vector{S}) where {T<:Real, S<:Real} = Categorical(Vector{T}(p))
convert(::Type{Categorical{T}}, d::Categorical{S}) where {T<:Real, S<:Real} = Categorical(Vector{T}(probs(d)))
convert(::Type{Categorical{P,Ps}}, x::AbstractVector{<:Real}) where {
P<:Real,Ps<:AbstractVector{P}} = Categorical{P,Ps}(Ps(x))

# convert(::Type{Categorical{P,Ps}}, d::Categorical) where {
# P<:Real,Ps<:AbstractVector{P}} = Categorical{P,Ps}(Ps(probs(d)))

### Parameters

ncategories(d::Categorical) = support(d).stop
params(d::Categorical) = (probs(d),)
params(d::Categorical{P,Ps}) where {P<:Real, Ps<:AbstractVector{P}} = (probs(d),)
@inline partype(d::Categorical{T}) where {T<:Real} = T

### Statistics
Expand Down Expand Up @@ -112,7 +117,8 @@ end

# sampling

sampler(d::Categorical) = AliasTable(d.p)
sampler(d::Categorical{P,Ps}) where {P<:Real,Ps<:AbstractVector{P}} =
AliasTable(probs(d))


### sufficient statistics
Expand Down
51 changes: 28 additions & 23 deletions src/univariate/discrete/discretenonparametric.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,28 +17,31 @@ External links

* [Probability mass function on Wikipedia](http://en.wikipedia.org/wiki/Probability_mass_function)
"""
struct DiscreteNonParametric{T<:Real,P<:Real,S<:AbstractVector{T}} <: DiscreteUnivariateDistribution
support::S
p::Vector{P}
struct DiscreteNonParametric{T<:Real,P<:Real,Ts<:AbstractVector{T},Ps<:AbstractVector{P}} <: DiscreteUnivariateDistribution
support::Ts
p::Ps

DiscreteNonParametric{T,P,S}(vs::S, ps::Vector{P}, ::NoArgCheck) where {T<:Real,P<:Real,S<:AbstractVector{T}} =
new{T,P,S}(vs, ps)
DiscreteNonParametric{T,P,Ts,Ps}(vs::Ts, ps::Ps, ::NoArgCheck) where {
T<:Real,P<:Real,Ts<:AbstractVector{T},Ps<:AbstractVector{P}} =
new{T,P,Ts,Ps}(vs, ps)

function DiscreteNonParametric{T,P,S}(vs::S, ps::Vector{P}) where {T<:Real,P<:Real,S<:AbstractVector{T}}
function DiscreteNonParametric{T,P,Ts,Ps}(vs::Ts, ps::Ps) where {
T<:Real,P<:Real,Ts<:AbstractVector{T},Ps<:AbstractVector{P}}
@check_args(DiscreteNonParametric, length(vs) == length(ps))
@check_args(DiscreteNonParametric, isprobvec(ps))
@check_args(DiscreteNonParametric, allunique(vs))
sort_order = sortperm(vs)
new{T,P,S}(vs[sort_order], ps[sort_order])
new{T,P,Ts,Ps}(vs[sort_order], ps[sort_order])
end
end

DiscreteNonParametric(vs::S, ps::Vector{P}) where {T<:Real,P<:Real,S<:AbstractVector{T}} =
DiscreteNonParametric{T,P,S}(vs, ps)
DiscreteNonParametric(vs::Ts, ps::Ps) where {
T<:Real,P<:Real,Ts<:AbstractVector{T},Ps<:AbstractVector{P}} =
DiscreteNonParametric{T,P,Ts,Ps}(vs, ps)

# Conversion
convert(::Type{DiscreteNonParametric{T,P,S}}, d::DiscreteNonParametric) where {T,P,S} =
DiscreteNonParametric{T,P,S}(S(support(d)), Vector{P}(probs(d)), NoArgCheck())
convert(::Type{DiscreteNonParametric{T,P,Ts,Ps}}, d::DiscreteNonParametric) where {T,P,Ts,Ps} =
DiscreteNonParametric{T,P,Ts,Ps}(Ts(support(d)), Ps(probs(d)), NoArgCheck())

# Accessors
params(d::DiscreteNonParametric) = (d.support, d.p)
Expand Down Expand Up @@ -110,9 +113,8 @@ function quantile(d::DiscreteNonParametric, q::Real)
x[i]
end


minimum(d::DiscreteNonParametric) = support(d)[1]
maximum(d::DiscreteNonParametric) = support(d)[end]
minimum(d::DiscreteNonParametric) = first(support(d))
maximum(d::DiscreteNonParametric) = last(support(d))
insupport(d::DiscreteNonParametric, x::Real) =
length(searchsorted(support(d), x)) > 0

Expand Down Expand Up @@ -205,9 +207,10 @@ end

# Sufficient statistics

struct DiscreteNonParametricStats{T<:Real,P<:AbstractFloat,S<:AbstractVector{T}} <: SufficientStats
support::S
freq::Vector{P}
struct DiscreteNonParametricStats{T<:Real,W<:Real,Ts<:AbstractVector{T},
Ws<:AbstractVector{W}} <: SufficientStats
support::Ts
freq::Ws
end

function suffstats(::Type{DiscreteNonParametric}, x::AbstractArray{T}) where {T<:Real}
Expand All @@ -217,7 +220,7 @@ function suffstats(::Type{DiscreteNonParametric}, x::AbstractArray{T}) where {T<

n = 1
vs = Vector{T}(undef,N)
ps = zeros(N)
ps = zeros(Float64, N)
x = sort(vec(x))

vs[1] = x[1]
Expand All @@ -240,16 +243,17 @@ function suffstats(::Type{DiscreteNonParametric}, x::AbstractArray{T}) where {T<

end

function suffstats(::Type{DiscreteNonParametric}, x::AbstractArray{T}, w::AbstractArray{P}) where {T<:Real,P<:AbstractFloat}
function suffstats(::Type{DiscreteNonParametric}, x::AbstractArray{T},
w::AbstractArray{W}) where {T<:Real,W<:Real}

@check_args(DiscreteNonParametric, length(x) == length(w))

N = length(x)
N == 0 && return DiscreteNonParametricStats(T[], P[])
N == 0 && return DiscreteNonParametricStats(T[], W[])

n = 1
vs = Vector{T}(undef, N)
ps = zeros(P, N)
ps = zeros(W, N)

xorder = sortperm(vec(x))
x = vec(x)[xorder]
Expand Down Expand Up @@ -278,5 +282,6 @@ end

# # Model fitting

fit_mle(::Type{DiscreteNonParametric}, ss::DiscreteNonParametricStats{T,P,S}) where {T,P,S} =
DiscreteNonParametric{T,P,S}(ss.support, pnormalize!(copy(ss.freq)), NoArgCheck())
fit_mle(::Type{DiscreteNonParametric},
ss::DiscreteNonParametricStats{T,W,Ts,Ws}) where {T,W,Ts,Ws} =
DiscreteNonParametric{T,W,Ts,Ws}(ss.support, pnormalize!(copy(ss.freq)), NoArgCheck())
12 changes: 1 addition & 11 deletions src/utils.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -79,17 +79,7 @@ end

isprobvec(p::Vector{T}) where {T<:Real} = allnonneg(p) && isapprox(sum(p), one(T))

function pnormalize!(v::AbstractVector{T}) where T<:AbstractFloat
s = 0.
n = length(v)
for i = 1:n
@inbounds s += v[i]
end
for i = 1:n
@inbounds v[i] /= s
end
v
end
pnormalize!(v::AbstractVector{<:Real}) = (v ./= sum(v); v)

add!(x::AbstractArray, y::AbstractVector) = broadcast!(+, x, x, y)
add!(x::AbstractVecOrMat, y::ZeroVector) = x
Expand Down
4 changes: 2 additions & 2 deletions test/categorical.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -52,5 +52,5 @@ println(" testing $d as Categorical")
p = ones(10^6) * 1.0e-6
@test Distributions.isprobvec(p)

@test typeof(convert(Categorical{Float32}, d)) == Categorical{Float32}
@test typeof(convert(Categorical{Float32}, d.p)) == Categorical{Float32}
@test typeof(convert(Categorical{Float32,Vector{Float32}}, d)) == Categorical{Float32,Vector{Float32}}
@test typeof(convert(Categorical{Float32,Vector{Float32}}, d.p)) == Categorical{Float32,Vector{Float32}}