bartorch.optim.EulerMaruyama

class bartorch.optim.EulerMaruyama(regularizers=None, *, step, maxiter=30, eigen=False, cclambda=0.0, precond=None, sampler_precond=None, sampler_precond_diag=0.0, sampler_precond_tol=0.0, sampler_precond_maxiter=10)

BART’s Euler-Maruyama iteration (pics --eulermaruyama).

Parameters:

• regularizers (Regularizer or iterable of Regularizer, optional) – Terms from bartorch.prox.

• step (float) – Step size (pics -s). Required: pics supplies no default for this iteration.

• maxiter (int)

• eigen (bool) – Scale the step by the largest eigenvalue of the normal operator, estimated with 30 power iterations (pics -e).

• cclambda (float) – Weight of an identity added to the normal operator (pics -q).

__init__(regularizers=None, *, step, maxiter=30, eigen=False, cclambda=0.0, precond=None, sampler_precond=None, sampler_precond_diag=0.0, sampler_precond_tol=0.0, sampler_precond_maxiter=10)

Methods

__init__([regularizers, maxiter, eigen, ...])
fixed_point(image_shape, *[, trainable])	This solver as a deep-equilibrium model: the step's fixed point.
in_library(y, A[, x0])	Solve with BART's own loop, without crossing back into Python.
unrolled(image_shape, *[, trainable])	This solver as a network of maxiter steps, trained end to end.