update name again

Project.toml

@@ -1,4 +1,4 @@
-name = "SequentialTransportMaps"
+name = "SequentialMeasureTransport"
 uuid = "880b6c9a-c7f7-4f83-b244-c5d4e47b6fa9"
 authors = ["Benjamin Zanger <[email protected]>"]
 version = "0.1.0"

README.md

@@ -1,4 +1,4 @@
-# SequentialTransportMaps.jl
+# SequentialMeasureTransport.jl
 
 Code of the paper [Seqential Transport maps using SoS densities and $\alpha$-divergences](https://arxiv.org/abs/2402.17943)
 

src/MCMC/mcmc.jl

@@ -4,8 +4,8 @@ Implements MCMC based on the `AbstractMCMC` interface using the
 sampler from `Sampler` in this package.
 """
 
-using ..SequentialTransportMaps
-using ..SequentialTransportMaps: AbstractSampler
+using ..SequentialMeasureTransport
+using ..SequentialMeasureTransport: AbstractSampler
 import AbstractMCMC
 import MCMCChains
 using Distributions
@@ -29,7 +29,7 @@ end
 function (a::MCMCModel{F})(x::Vector{T}, 
                 sampler::AbstractSampler{<:Any, <:Any, T}
             ) where {F, T<:Number}
-    return SequentialTransportMaps.pullback(sampler, x->a.model(x))(x)
+    return SequentialMeasureTransport.pullback(sampler, x->a.model(x))(x)
 end
 
 function proposal(sampler::MCMCSampler{T}, x::Vector{T}) where {T}
@@ -42,7 +42,7 @@ function AbstractMCMC.step(rng::Random.AbstractRNG,
         model::MCMCModel,
         samp::MCMCSampler{T};
         kwargs...) where {T <: Number}
-    next_θ = SequentialTransportMaps.sample(samp.sampler)
+    next_θ = SequentialMeasureTransport.sample(samp.sampler)
     return next_θ, next_θ
 end
 function AbstractMCMC.step(rng::Random.AbstractRNG,

src/PSDModels/models.jl

@@ -1,7 +1,7 @@
 include("kernel/PSDModelKernel.jl")
 include("feature_map/PSDModelFM.jl")
 
-# optimization algorithms acting on SequentialTransportMaps
+# optimization algorithms acting on SequentialMeasureTransport
 include("optimization.jl")
 
 function PSDModel(k::Kernel, X::PSDDataVector{T}; kwargs...) where {T<:Number}

src/Samplers/bridging/bridging_densities.jl

@@ -3,7 +3,7 @@ module BridgingDensities
 abstract type BridgingDensity{d, T} end
 
 using DifferentialEquations
-using ..SequentialTransportMaps: PSDDataVector, PSDdata
+using ..SequentialMeasureTransport: PSDDataVector, PSDdata
 
 include("algebraic.jl")
 include("diffusion.jl")

src/Samplers/reference_maps/algebraic.jl

@@ -9,15 +9,15 @@ struct AlgebraicReference{d, T} <: ReferenceMap{d, T}
 end
 
 
-function SequentialTransportMaps.pushforward(
+function SequentialMeasureTransport.pushforward(
         m::AlgebraicReference{<:Any, T}, 
         x::PSDdata{T}
     ) where {T<:Number}
     return ((x./sqrt.(1 .+ x.^2)).+1.0)/2.0
 end
 
 
-function SequentialTransportMaps.pullback(
+function SequentialMeasureTransport.pullback(
         m::AlgebraicReference{<:Any, T}, 
         u::PSDdata{T}
     ) where {T<:Number}
@@ -26,15 +26,15 @@ function SequentialTransportMaps.pullback(
 end
 
 
-function SequentialTransportMaps.Jacobian(
+function SequentialMeasureTransport.Jacobian(
         m::AlgebraicReference{<:Any, T}, 
         x::PSDdata{T}
     ) where {T<:Number}
     return mapreduce(xi->0.5/(1+xi^2)^(3/2), *, x)
 end
 
 
-function SequentialTransportMaps.inverse_Jacobian(
+function SequentialMeasureTransport.inverse_Jacobian(
         mapping::AlgebraicReference{<:Any, T}, 
         u::PSDdata{T}
     ) where {T<:Number}

src/Samplers/reference_maps/gaussian.jl

@@ -25,31 +25,31 @@ function sample_reference(map::GaussianReference{d, T}, n::Int) where {d, T<:Num
     eachcol(randn(T, d, n) * map.σ)
 end
 
-function SequentialTransportMaps.pushforward(
+function SequentialMeasureTransport.pushforward(
         m::GaussianReference{d, T}, 
         x::PSDdata{T}
     ) where {d, T<:Number}
     return 0.5 * (1 .+ erf.(x ./ (m.σ * sqrt(2))))
 end
 
 
-function SequentialTransportMaps.pullback(
+function SequentialMeasureTransport.pullback(
         m::GaussianReference{d, T}, 
         u::PSDdata{T}
     ) where {d, T<:Number}
     return sqrt(2) * m.σ * erfcinv.(2.0 .- 2*u)
 end
 
 
-function SequentialTransportMaps.Jacobian(
+function SequentialMeasureTransport.Jacobian(
         m::GaussianReference{d, T}, 
         x::PSDdata{T}
     ) where {d, T<:Number}
     return mapreduce(xi->Distributions.pdf(Distributions.Normal(0, m.σ), xi), *, x)
 end
 
 
-function SequentialTransportMaps.inverse_Jacobian(
+function SequentialMeasureTransport.inverse_Jacobian(
         mapping::GaussianReference{d, T}, 
         u::PSDdata{T}
     ) where {d, T<:Number}

src/Samplers/reference_maps/reference_maps.jl

@@ -1,7 +1,7 @@
 module ReferenceMaps
 
-import ..SequentialTransportMaps
-using ..SequentialTransportMaps: PSDModelOrthonormal, domain_interval_left, domain_interval_right,
+import ..SequentialMeasureTransport
+using ..SequentialMeasureTransport: PSDModelOrthonormal, domain_interval_left, domain_interval_right,
                    PSDdata, Mapping, pullback, pushforward, Jacobian, inverse_Jacobian
 
 using SpecialFunctions: erf, erfcinv
@@ -44,7 +44,7 @@ end
 
 Pushes forward a vector of a reference distribution to the uniform distribution.
 """
-function SequentialTransportMaps.pushforward(
+function SequentialMeasureTransport.pushforward(
         mapping::ReferenceMap{d, T}, 
         x::PSDdata{T}
     ) where {d, T<:Number}
@@ -56,7 +56,7 @@ end
 
 Pulls back a vector of the uniform distribution to the reference distribution.
 """
-function SequentialTransportMaps.pullback(
+function SequentialMeasureTransport.pullback(
         mapping::ReferenceMap{d, T}, 
         u::PSDdata{T}
     ) where {d, T<:Number}
@@ -68,7 +68,7 @@ end
 
 Computes the Jacobian of the mapping at the point u.
 """
-function SequentialTransportMaps.Jacobian(
+function SequentialMeasureTransport.Jacobian(
         mapping::ReferenceMap{d, T}, 
         x::PSDdata{T}
     ) where {d, T<:Number}
@@ -80,7 +80,7 @@ inverse_Jacobian(mapping, u)
 
 Computes the inverse Jacobian of the mapping at the point x.
 """
-function SequentialTransportMaps.inverse_Jacobian(
+function SequentialMeasureTransport.inverse_Jacobian(
         mapping::ReferenceMap{d, T}, 
         u::PSDdata{T}
     ) where {d, T<:Number}

src/Samplers/reference_maps/scaling.jl

@@ -27,7 +27,7 @@ end
 
 ## Interface implementation
 
-function SequentialTransportMaps.pushforward(
+function SequentialMeasureTransport.pushforward(
         m::ScalingReference{d, T}, 
         x::PSDdata{T}
     ) where {d, T<:Number}
@@ -36,7 +36,7 @@ function SequentialTransportMaps.pushforward(
 end
 
 
-function SequentialTransportMaps.pullback(
+function SequentialMeasureTransport.pullback(
         m::ScalingReference{d, T}, 
         u::PSDdata{T}
     ) where {d, T<:Number}
@@ -45,7 +45,7 @@ function SequentialTransportMaps.pullback(
 end
 
 
-function SequentialTransportMaps.Jacobian(
+function SequentialMeasureTransport.Jacobian(
         mapping::ScalingReference{d, T}, 
         x::PSDdata{T}
     ) where {d, T<:Number}
@@ -57,7 +57,7 @@ function SequentialTransportMaps.Jacobian(
 end
 
 
-function SequentialTransportMaps.inverse_Jacobian(
+function SequentialMeasureTransport.inverse_Jacobian(
         mapping::ScalingReference{d, T}, 
         u::PSDdata{T}
     ) where {d, T<:Number}

src/SequentialTransportMaps.jl → src/SequentialMeasureTransport.jl

@@ -1,4 +1,4 @@
-module SequentialTransportMaps
+module SequentialMeasureTransport
 
 using LinearAlgebra, SparseArrays
 using KernelFunctions: Kernel, kernelmatrix

src/statistics.jl

@@ -1,14 +1,14 @@
 module Statistics
 
-using ..SequentialTransportMaps
-using ..SequentialTransportMaps: PSDDataVector
-using ..SequentialTransportMaps: PSDModelOrthonormal
-using ..SequentialTransportMaps: CondSampler
-using ..SequentialTransportMaps: domain_interval_left, domain_interval_right
-using ..SequentialTransportMaps: greedy_IRLS
-using ..SequentialTransportMaps: _ML_JuMP!
-using ..SequentialTransportMaps: _KL_JuMP!
-using ..SequentialTransportMaps: _α_divergence_JuMP!
+using ..SequentialMeasureTransport
+using ..SequentialMeasureTransport: PSDDataVector
+using ..SequentialMeasureTransport: PSDModelOrthonormal
+using ..SequentialMeasureTransport: CondSampler
+using ..SequentialMeasureTransport: domain_interval_left, domain_interval_right
+using ..SequentialMeasureTransport: greedy_IRLS
+using ..SequentialMeasureTransport: _ML_JuMP!
+using ..SequentialMeasureTransport: _KL_JuMP!
+using ..SequentialMeasureTransport: _α_divergence_JuMP!
 using LinearAlgebra
 using FastGaussQuadrature: gausslegendre
 using Distributions: pdf

test/PSDKernel_test.jl

@@ -160,7 +160,7 @@ end
     model = PSDModel(X, Y, k; solver=:gradient_descent)
 
     @inline interval(x) = 0..x[1]
-    int_vec = SequentialTransportMaps.integrate.(Ref(model), interval.(X))
+    int_vec = SequentialMeasureTransport.integrate.(Ref(model), interval.(X))
     @test norm(int_vec - f_int.(X)) < 1e-1
 end
 
@@ -195,7 +195,7 @@ end
 
     X2 = Float64[4, 5, 6]
     X2 = [[x] for x in X2]
-    model2 = SequentialTransportMaps.add_support(model, X2)
+    model2 = SequentialMeasureTransport.add_support(model, X2)
 
     for i in rand(20) * 4
         @test model(i) ≈ model2(i)

test/PSDPolynomial_test.jl

@@ -146,7 +146,7 @@ end
         normalize_orth_measure!(model)
         @test tr(model.B) ≈ 1.0
 
-        SequentialTransportMaps.normalize!(model)
+        SequentialMeasureTransport.normalize!(model)
         @test tr(model.B) ≈ 1.0
     end
 
@@ -162,7 +162,7 @@ end
             normalize_orth_measure!(model)
             @test tr(model.B) ≈ 1.0
 
-            SequentialTransportMaps.normalize!(model)
+            SequentialMeasureTransport.normalize!(model)
             @test tr(model.B) ≈ 1.0
         end
     end
@@ -177,7 +177,7 @@ end
         Y = f.(X)
 
         fit!(model, X, Y, maxit=2000)
-        model_perm = SequentialTransportMaps.permute_indices(model, [2,1])
+        model_perm = SequentialMeasureTransport.permute_indices(model, [2,1])
         for i=1:100
             x = (rand(2) .- 0.5) * 30.0
             @test model_perm(x) == model([x[2], x[1]])
@@ -194,7 +194,7 @@ end
         Y = f.(X)
 
         fit!(model, X, Y, maxit=2000)
-        model_poly = SequentialTransportMaps.compile(model)
+        model_poly = SequentialMeasureTransport.compile(model)
         for i=1:100
             x = (rand(2) .- 0.5) * 30.0
             @test model_poly(x) ≈ model(x)
@@ -213,11 +213,11 @@ end
         fit!(model, X, Y)
 
         int_model = integral(model, 1; C=0.0)
-        int_model_compiled = SequentialTransportMaps.compiled_integral(model, 1; C=0.0)
+        int_model_compiled = SequentialMeasureTransport.compiled_integral(model, 1; C=0.0)
         @test all(int_model.(X) .≈ int_model_compiled.(X))
 
         int_model2 = integral(model, 2; C=0.0)
-        int_model_compiled2 = SequentialTransportMaps.compiled_integral(model, 2; C=0.0)
+        int_model_compiled2 = SequentialMeasureTransport.compiled_integral(model, 2; C=0.0)
         @test all(int_model2.(X) .≈ int_model_compiled2.(X))
     end
 
@@ -233,13 +233,13 @@ end
         fit!(model, X, Y)
 
         int_model = integral(model, 1)
-        int_model_compiled = SequentialTransportMaps.compiled_integral(model, 1)
+        int_model_compiled = SequentialMeasureTransport.compiled_integral(model, 1)
         for x in X
             @test isapprox(int_model(x), int_model_compiled(x), atol=1e-10)
         end
 
         int_model2 = integral(model, 2)
-        int_model_compiled2 = SequentialTransportMaps.compiled_integral(model, 2)
+        int_model_compiled2 = SequentialMeasureTransport.compiled_integral(model, 2)
         for x in X
             @test isapprox(int_model2(x), int_model_compiled2(x), atol=1e-10)
         end

test/Sampler/SR_ML_test.jl

@@ -11,8 +11,8 @@
         samples = [[x] for x in samples]
         model = PSDModel(Legendre(0.0..1.0), :downward_closed, 3)
         bridge = DiffusionBrigdingDensity{1}(x->1.0, [1.0, 0.5, 0.25, 0.1, 0.05, 0.0], 2.0)
-        ref_map = SequentialTransportMaps.ReferenceMaps.GaussianReference{1, Float64}(2.0)
-        sra = SequentialTransportMaps.SelfReinforced_ML_estimation(
+        ref_map = SequentialMeasureTransport.ReferenceMaps.GaussianReference{1, Float64}(2.0)
+        sra = SequentialMeasureTransport.SelfReinforced_ML_estimation(
             samples,
             model,
             bridge,
@@ -38,8 +38,8 @@ end
         X_train = [[x] for x in X_train]
         X_val = [[x] for x in X_val]
         model = PSDModel(Legendre(0.0..1.0), :downward_closed, 5)
-        ref_map = SequentialTransportMaps.ReferenceMaps.AlgebraicReference{1, Float64}()
-        sra = SequentialTransportMaps.Adaptive_Self_reinforced_ML_estimation(
+        ref_map = SequentialMeasureTransport.ReferenceMaps.AlgebraicReference{1, Float64}()
+        sra = SequentialMeasureTransport.Adaptive_Self_reinforced_ML_estimation(
             X_train,
             X_val,
             model,