sheap.MasterSampler.Samplers.PseudoMonteCarloSampler module

Pseudo Monte Carlo Sampler

Note

the docs require update now.

?

class PseudoMonteCarloSampler(estimator, dtype=jax.numpy.float32)[source]

Bases: object

Laplace (pseudo Monte Carlo) sampler in PHYSICAL space, robust and batched. ?

Parameters:

estimator (Any)

sample_params(num_samples, key_seed=0, cov_phys=None, residuals_fn_phys=None, eps=1e-08, summarize=True)[source]
Returns:

  • “samples_raw”: (S, D) or (N, S, D)

  • ”samples_phys”: (S, D) or (N, S, D) (always included)

  • ”products”: postprocess_fn(samples_phys) if provided

Return type:

dict with keys

Parameters:

  • num_samples (int)

  • key_seed (int)

  • cov_phys (jax.numpy.ndarray | None)

  • residuals_fn_phys (Callable[[jax.numpy.ndarray], jax.numpy.ndarray] | None)

  • eps (float)

  • summarize (bool)