Merge pull request #153 from Julia-Tempering/banana-scale

Add scale parameter to banana distribution
Julia-Tempering · Oct 8, 2023 · 02c1c14 · 02c1c14
2 parents 9d7e6e9 + 3a7a426
commit 02c1c14
Show file tree

Hide file tree

Showing 4 changed files with 6 additions and 5 deletions.
diff --git a/examples/stan/banana.stan b/examples/stan/banana.stan
@@ -5,6 +5,7 @@
 // y_{1:d}|x ~ N(x^2, s_b) // s_b = sqrt(inv(2b)), b = 5    (50 easier)
 data {
     int<lower=1> dim;
+    real<lower=0> scale;
 }
 transformed data {
     real a, b, s_a, s_b;
@@ -19,5 +20,5 @@ parameters {
 }
 model {
     x ~ normal(0, s_a);
-    y ~ normal(square(x), s_b);
+    y ~ normal(square(x), scale*s_b);
 }
diff --git a/src/targets/toy_stan_target.jl b/src/targets/toy_stan_target.jl
@@ -29,10 +29,10 @@ stan_bernoulli(y = [0,1,0,0,0,0,0,0,0,1]) =
         json(; y, N = length(y))
     )
 
-stan_banana(dim = 9) = 
+stan_banana(dim = 9, scale = 1.0) = 
     StanLogPotential(
         stan_example_path("banana.stan"), 
-        json(; dim)
+        json(; dim, scale)
     )
 
 observed_range_squared(x) = (maximum(x) - minimum(x))^2

diff --git a/test/test_AAPS.jl b/test/test_AAPS.jl
@@ -2,7 +2,7 @@ using MCMCChains
 if !is_windows_in_CI()
     @testset "AAPS" begin
         pt = pigeons(; 
-            target = Pigeons.stan_banana(1), 
+            target = Pigeons.stan_banana(1, 1.0), 
             explorer = AAPS(step_size = 2. ^(-4)), 
             n_chains = 1, n_rounds = 12, record = [traces])
         @test abs(23-minimum(ess(Chains(sample_array(pt))).nt.ess)) < 1            

diff --git a/test/test_stan.jl b/test/test_stan.jl
@@ -2,7 +2,7 @@ if !is_windows_in_CI()
 @testset "Stan examples" begin
     pigeons(target = Pigeons.stan_eight_schools(true), n_rounds = 2, n_chains = 2)
     pigeons(target = Pigeons.stan_eight_schools(false), n_rounds = 2, n_chains = 2)
-    pigeons(target = Pigeons.stan_banana(1), record = [online], n_chains = 1, n_rounds = 5, explorer = SliceSampler())
+    pigeons(target = Pigeons.stan_banana(1, 1.0), record = [online], n_chains = 1, n_rounds = 5, explorer = SliceSampler())
 
     # some examples where an error is interpreted as -Inf:
     pigeons(target = Pigeons.stan_funnel(1), record = [online], n_chains = 1, n_rounds = 5, explorer = SliceSampler())