Add PMI profile drawing

workflows/Mixtures/cool_tasks.smk

@@ -36,7 +36,7 @@ x_sampler = fine.FineSampler(x_dist)
 # The fence distribution
 n_components = 12
 
-base_dist = fine.mixture(
+fence_base_dist = fine.mixture(
     proportions=jnp.ones(n_components) / n_components,
     components=[
         fine.MultivariateNormalDistribution(
@@ -46,7 +46,7 @@ base_dist = fine.mixture(
         ) for x in range(n_components)
     ]
 )
-base_sampler = fine.FineSampler(base_dist)
+base_sampler = fine.FineSampler(fence_base_dist)
 fence_aux_sampler = bmi.samplers.TransformedSampler(
     base_sampler,
     transform_x=lambda x: x + jnp.array([5., 0.]) * jnp.sin(3 * x[1]),
@@ -192,7 +192,8 @@ rule all:
     input:
         'cool_tasks.pdf',
         'results.csv',
-        'results.pdf'
+        'results.pdf',
+        'profiles.pdf'
 
 rule plot_distributions:
     output: "cool_tasks.pdf"
@@ -235,6 +236,23 @@ rule plot_distributions:
 
         fig.savefig(str(output))
 
+rule plot_pmi_profiles:
+    output: "profiles.pdf"
+    run:
+        fig, axs = subplots_from_axsize(1, 4, axsize=(4, 3))
+        dists = [x_dist, ai_dist, fence_base_dist, balls_mixt]
+        tasks_official = ['X', 'AI', 'Waves', 'Galaxy']
+        for dist, task_name, ax in zip(dists, tasks_official, axs):
+            import jax
+            key = jax.random.PRNGKey(1024)
+            pmi_values = fine.pmi_profile(key=key, dist=dist, n=100_000)
+            bins = np.linspace(-5, 5, 101)
+            ax.hist(pmi_values, bins=bins, density=True, alpha=0.5)
+            ax.set_xlabel(task_name)
+        axs[0].set_ylabel("Density")
+        fig.savefig(str(output))
+
+
 rule plot_results:
     output: 'results.pdf'
     input: 'results.csv'