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.

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

Multithreaded, lighter sampling #114

Merged
merged 1 commit into from
Feb 28, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
2 changes: 1 addition & 1 deletion src/MatrixProductBP.jl
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,7 @@ import Statistics: mean, std
import Unzip: unzip
import StatsBase: weights, proportions
import LogExpFunctions: logistic, logsumexp
import .Threads: SpinLock, lock, unlock
import .Threads: SpinLock, lock, unlock, @threads
import Lazy: @forward
import CavityTools: cavity
import LogarithmicNumbers: ULogarithmic
Expand Down
24 changes: 13 additions & 11 deletions src/sampling.jl
Original file line number Diff line number Diff line change
Expand Up @@ -3,25 +3,25 @@ import CavityTools: ExponentialQueue

# as in https://doi.org/10.1103/PhysRevLett.114.248701
# draw samples from the prior and weight them with their likelihood
struct SoftMarginSampler{B<:MPBP, F<:Real}
struct SoftMarginSampler{B<:MPBP,F<:Real,TI<:Integer}
bp :: B
X :: Vector{Matrix{Int}}
X :: Vector{Matrix{TI}}
w :: Vector{F}

function SoftMarginSampler(bp::TBP,
X::Vector{Matrix{Int}}, w::Vector{F}) where{TBP<:MPBP, F}
X::Vector{Matrix{TI}}, w::Vector{F}) where{TBP<:MPBP,F,TI<:Integer}

N = nv(bp.g); T = getT(bp)
@assert length(X) == length(w)
@assert all(≥(0), w)
@assert all(x -> size(x) == (N, T+1), X)

new{TBP, F}(bp, X, w)
new{TBP,F,TI}(bp, X, w)
end
end

function SoftMarginSampler(bp::MPBP)
X = Matrix{Int}[]
X = Matrix{UInt8}[]
w = zeros(ULogarithmic, 0)
SoftMarginSampler(bp, X, w)
end
Expand Down Expand Up @@ -72,7 +72,7 @@ function sample!(sms::SoftMarginSampler, nsamples::Integer;
X = [zeros(Int, N, T+1) for _ in 1:nsamples]
p⁰ = [ϕᵢ[1] ./ sum(ϕᵢ[1]) for ϕᵢ in sms.bp.ϕ]
w = zeros(ULogarithmic, nsamples)
for n in 1:nsamples
@threads for n in 1:nsamples
_, w[n] = onesample!(X[n], sms.bp; p⁰, rng)
next!(prog)
end
Expand All @@ -98,9 +98,11 @@ function marginals(sms::SoftMarginSampler; showprogress::Bool=true, sites=vertic
prog = Progress(N, desc="Marginals from Soft Margin"; dt=showprogress ? 0.1 : Inf)
is_free = is_free_dynamics(sms.bp)

for (a,i) in pairs(sites)
@threads for a in eachindex(sites)
i = sites[a]
x = x = zeros(Int, length(X))
for t in 1:T+1
x = [xx[i, t] for xx in X]
x .= [xx[i, t] for xx in X]
mit_avg = is_free ? proportions(x, nstates(bp,i)) : proportions(x, nstates(bp,i), wv)
mit_var = mit_avg .* (1 .- mit_avg) ./ nsamples
marg[a][t] .= mit_avg .± sqrt.( mit_var )
Expand All @@ -126,11 +128,11 @@ function pair_marginals(sms::SoftMarginSampler; showprogress::Bool=true)
@assert all(>=(0), w)
wv = weights(w)
nsamples = length(X)
prog = Progress(E, desc="Marginals from Soft Margin"; dt=showprogress ? 0.1 : Inf)
x = zeros(Int, length(X))
prog = Progress(E, desc="Pair marginals from Soft Margin"; dt=showprogress ? 0.1 : Inf)

for (i,j,id) in edges(g)
@threads for (i,j,id) in collect(edges(g))
linear = LinearIndices((1:nstates(bp,i), 1:nstates(bp,j)))
x = zeros(Int, length(X))
for t in 1:T+1
x .= [linear[xx[i, t],xx[j,t]] for xx in X]
mijt_avg_linear = proportions(x, nstates(bp,i)*nstates(bp,j), wv)
Expand Down
Loading