Benchmark amortized simulation-based inference (SBI) algorithms.
Datasets:
- train/test: parameter-observation pairs from the prior-simulator
joint distribution :math:`p(\theta, x)=p(\theta)p(x|\theta)`.
- reference (optional): observations math:`x_ref` and corresponding
samples from the reference posterior :math:`p(\theta|x_ref)`
(if available).
Solvers: amortized SBI algorithms trained on the joint to approximate
the posterior :math:`p(\theta|x)` for any observation :math:`x`.
Metrics:
- expected negative log likelihood (NLL) on test (stopping criterion)
and train datasets.
- C2ST, EMD, MMD on reference dataset (optional).