isofit.configs.sections.inversion_config

class InversionConfig(sub_configdic=None)

Bases: BaseConfigSection

Inversion configuration.

windows

inversion retrieval windows to operate over.

Type:

List[List[float]]

cressie_map_confidence

N. Cressie [ASA 2018] suggests an alternate definition of S_hat for more statistically-consistent posterior confidence estimation, this flag runs in this mode

Type:

bool

mcmc

MCMC parameters, only used if mode = mcmc.

integration_grid

Grid of inversion points to execute if mode=’grid’. Either fixed, or starting points, depending on self.fixed_inversion_grid

priors_in_initial_guess

Boolean to inidicate the use of surface priors outside of the inversion windows during the intial guess.

inversion_grid_as_preseed

Parameter indicating whether to treat the inversion grid as: (True) - a series of seeds for the optimization (variable by the optimization algorithm). (False) - a set of fixed points (not variable by the optimization algorithm)

least_squares_params

Least squares parameters for core inversion solve. See https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.least_squares.html for details.

class McmcConfig(sub_configdic=None)

Bases: BaseConfigSection

MCMC inversion configuration.

iterations

Number of MCMC iterations to run.

Type:

int

class LeastSquaresConfig(sub_configdic=None)

Bases: BaseConfigSection

Least squares config parameters.

method

Optimzation method to use. Default ‘trf’.

Type:

str

max_nfev

Maximum number of function evaluations before the termination. If None (default), the value is chosen automatically. Default 20.

Type:

int

xtol

Tolerance for termination by the change of the independent variables. Default is None, which disables termination from this criteria.

Type:

float

ftol

Tolerance for termination by the change of the cost function. Default is 0.01

Type:

float

gtol

Tolerance for termination by the norm of the gradient. Default is None, which disables termination from this criteria.

Type:

float

tr_solver

Method for solving trust-region subproblems, relevant only for ‘trf’ and ‘dogbox’ methods. Options are None, ‘exact’, or ‘lsmr’

Type:

str