benchmarks

tests/benchmarks.py

@@ -0,0 +1,28 @@
+from benchmarks.benchmarks_mchmc import *
+from sampling.sampler import Sampler
+from benchmarks import german_credit
+import pandas as pd
+
+
+
+def test():
+
+ names = ['Ill-Conditioned', 'Bi-Modal', 'Rosenbrock', "Neal's Funnel", 'German Credit', 'Stochastic Volatility']
+ targets = [IllConditionedGaussian(100, 100.0), BiModal(), Rosenbrock(), Funnel(), german_credit.Target(), StochasticVolatility()]
+
+ def ESS(target, num_samples):
+ sampler = Sampler(target, integrator= 'MN')
+ ess= jnp.average(sampler.sample(num_samples, 12, output= 'ess'))
+ return ess, sampler.L / np.sqrt(target.d), sampler.eps
+
+ num_samples= [30000, 100000, 300000, 100000, 100000, 10000]
+
+ results = np.array([ESS(targets[i], num_samples[i]) for i in range(len(names))])
+ df = pd.DataFrame({'Target ': names, 'ESS': results[:, 0], 'alpha': results[:, 1], 'eps': results[:, 2]})
+ print(df)
+
+
+
+
+test()
+