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

A more universal entropy calculation method for sampling based inference #151

Merged
merged 11 commits into from
Feb 23, 2021
Merged

A more universal entropy calculation method for sampling based inference #151

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
9da97e6
Avoid unnecessary matrix inversions
semihakbayrak Jan 23, 2021
85e7582
Reverted parametrization back to mean-variance.
ivan-bocharov Jan 28, 2021
f9a8a4b
Allow entropy calculation for smoothing
semihakbayrak Jan 28, 2021
5b49187
Initial tests for entropy calculation and new SampleList features
semihakbayrak Jan 29, 2021
ac77dbf
Renamed variables in SampleList tests.
ivan-bocharov Jan 29, 2021
fcba589
concatenateGaussianMV now can concatenate distributions of different …
ivan-bocharov Jan 29, 2021
8e08400
Fixed a typo in ruleSPNonlinearSInGX definition.
ivan-bocharov Jan 29, 2021
8e8a513
Folded logIntegrand initialization into a conditional branch.
ivan-bocharov Jan 29, 2021
5907b57
Add more comments about Entropy calculation. Test the new entropy cal…
semihakbayrak Feb 12, 2021
43ac935
Changes in convergence criteria of Laplace approximation
semihakbayrak Feb 18, 2021
376127d
Added DataStructures to the list of dependencies.
ivan-bocharov Feb 23, 2021
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
Initial tests for entropy calculation and new SampleList features
  • Loading branch information
@semihakbayrak
semihakbayrak committed Jan 29, 2021
commit 5b4918737314abd1f3b980dec774803bdf3cb94e
11 changes: 5 additions & 6 deletions src/factor_nodes/sample_list.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -132,10 +132,9 @@ isProper(dist::ProbabilityDistribution{V, SampleList}) where V<:VariateType = ab
y::ProbabilityDistribution{V, SampleList},
z::ProbabilityDistribution{V, SampleList}=ProbabilityDistribution(V, SampleList, s=[0.0], w=[1.0])) where V<:VariateType

logIntegrand = (samples) -> logPdf.([x], samples)
if haskey(y.params, :logintegrand)
logIntegrand(samples) = y.params[:logintegrand](samples) .+ logPdf.([x], samples)
else
logIntegrand(samples) = logPdf.([x], samples)
logIntegrand = (samples) -> y.params[:logintegrand](samples) .+ logPdf.([x], samples)
end

samples = y.params[:s]
Expand All @@ -156,11 +155,11 @@ isProper(dist::ProbabilityDistribution{V, SampleList}) where V<:VariateType = ab

z.params[:w] = weights
z.params[:s] = samples
z.params[:logintegrand] = (s) -> logIntegrand(s)
z.params[:logintegrand] = logIntegrand
if haskey(y.params, :logproposal) && haskey(y.params, :unnormalizedweights)
z.params[:unnormalizedweights] = w_raw_x.*y.params[:unnormalizedweights]
logProposal(s) = y.params[:logproposal](s)
z.params[:logproposal] = (s) -> logProposal(s)
logProposal = y.params[:logproposal]
z.params[:logproposal] = logProposal
# calculate entropy
H_y = log(sum(w_raw_x.*y.params[:unnormalizedweights])) - log(n_samples)
H_x = -sum( weights.*(logProposal(samples) + log.(y.params[:unnormalizedweights]) + log_samples_x) )
Expand Down
38 changes: 35 additions & 3 deletions test/factor_nodes/test_sample_list.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,8 +29,40 @@ end
end

@testset "prod!" begin
@test ProbabilityDistribution(Univariate, SampleList, s=[0.0, 1.0], w=[0.5, 0.5]) * ProbabilityDistribution(Univariate, GaussianMeanVariance, m=0.0, v=1.0) == ProbabilityDistribution(Univariate, SampleList, s=[0.0, 1.0], w=[0.6224593312018546, 0.37754066879814546])
@test ProbabilityDistribution(Multivariate, SampleList, s=[[0.0], [1.0]], w=[0.5, 0.5]) * ProbabilityDistribution(Multivariate, GaussianMeanVariance, m=[0.0], v=mat(1.0)) == ProbabilityDistribution(Multivariate, SampleList, s=[[0.0], [1.0]], w=[0.6224593312018546, 0.37754066879814546])
dist1 = ProbabilityDistribution(Univariate, GaussianMeanVariance, m=0.0, v=1.0)
dist2 = dist1 * ProbabilityDistribution(Univariate, SampleList, s=[0.0, 1.0], w=[0.5, 0.5])
dist3 = ProbabilityDistribution(Univariate, SampleList, s=[0.0, 1.0], w=[0.6224593312018546, 0.37754066879814546])
@test dist2.params[:s] == dist3.params[:s]
@test dist2.params[:w] == dist3.params[:w]
@test dist2.params[:logintegrand]([-0.7, 1.5]) == logPdf.([dist1],[-0.7, 1.5])
#@test ProbabilityDistribution(Multivariate, SampleList, s=[[0.0], [1.0]], w=[0.5, 0.5]) * ProbabilityDistribution(Multivariate, GaussianMeanVariance, m=[0.0], v=mat(1.0)) == ProbabilityDistribution(Multivariate, SampleList, s=[[0.0], [1.0]], w=[0.6224593312018546, 0.37754066879814546])
dist1 = ProbabilityDistribution(Multivariate, GaussianMeanVariance, m=[0.0], v=mat(1.0))
dist2 = ProbabilityDistribution(Multivariate, SampleList, s=[[0.0], [1.0]], w=[0.5, 0.5]) * dist1
dist3 = ProbabilityDistribution(Multivariate, SampleList, s=[[0.0], [1.0]], w=[0.6224593312018546, 0.37754066879814546])
@test dist2.params[:s] == dist3.params[:s]
@test dist2.params[:w] == dist3.params[:w]
@test dist2.params[:logintegrand]([[-0.7], [1.5]]) == logPdf.([dist1],[[-0.7], [1.5]])
end

@testset "Initialization of logproposal, logintegrand, unnormalizedweights" begin
dist1 = ProbabilityDistribution(Univariate, Beta, a=2.0, b=1.0)
dist2 = ProbabilityDistribution(Univariate, Gamma, a=2.0, b=1.0)
dist3 = ProbabilityDistribution(Univariate, SampleList, s=[0.0, 1.0], w=[0.5, 0.5])
dist4 = dist1*dist2
dist5 = dist2*dist3
dist6 = dist4*dist2

@test haskey(dist4.params, :logproposal)
@test haskey(dist4.params, :logintegrand)
@test haskey(dist4.params, :unnormalizedweights)

@test !haskey(dist5.params, :logproposal)
@test haskey(dist5.params, :logintegrand)
@test !haskey(dist5.params, :unnormalizedweights)

@test haskey(dist6.params, :logproposal)
@test haskey(dist6.params, :logintegrand)
@test haskey(dist6.params, :unnormalizedweights)
end

@testset "bootstrap" begin
Expand All @@ -48,7 +80,7 @@ end

p1 = ProbabilityDistribution(Multivariate, GaussianMeanVariance, m=[2.0], v=mat(0.0))
p2 = ProbabilityDistribution(MatrixVariate, SampleList, s=[mat(tiny)], w=[1.0])
@test isapprox(bootstrap(p1, p2)[1][1], 2.0, atol=1e-4)
@test isapprox(bootstrap(p1, p2)[1][1], 2.0, atol=1e-4)
end


Expand Down