magni.cs.reconstruction.sl0._config moduleΒΆ
Module providing configuration options for the magni.cs.reconstruction.sl0
subpackage.
See also
magni.cs.reconstruction._config.Configger
- The Configger class used
Notes
This module instantiates the Configger class provided by
magni.utils.config
. The configuration options are the following:
- epsilon : float
- The precision parameter used in centering (the default is 0.01).
- L : {‘geometric’, ‘fixed’}
- The method for selecting the maximum number of gradient descent iterations for each sigma.
- L_fixed : int
- The value used for L if using the ‘fixed’ method (the default is 2.0).
- L_geometric_ratio : float
- The common ratio used for the L update if using the ‘geometric’ method (the default is 2.0).
- L_geometric_start : float
- The starting value used for the L if using the ‘geometric’ method (the default is 2.0).
- mu : {‘step’, ‘fixed’}
- The method for selecting the relative step-size used in gradient descent iteration.
- mu_fixed : float
- The value used for mu if using the ‘fixed’ method (the default is 1.0).
- mu_step_end : float
- The value used for mu for the last iterations if using the ‘step’ method (the default is 1.5).
- mu_step_iteration : int
- The iteration where the value used for mu changes if using the ‘step’ method
- mu_step_start : float
- The value used for mu for the first iterations if using the ‘step’ method (the default is 0.001).
- precision_float : {np.float, np.float16, np.float32, np.float64, np.float128}
- The floating point precision used for the computations (the default is np.float64).
- sigma_geometric : float
- The common ratio used for the sigma update (the default is 0.7).
- sigma_start : {‘reciprocal’, ‘fixed’}
- The method for selecting the starting value of sigma.
- sigma_start_fixed : float
- The fixed factor multiplied onto the maximum coefficient of a least squares solution to obtain the value if using the ‘fixed’ method (the default is 2.0).
- sigma_start_reciprocal : float
- The constant in the factor \(\frac{1}{constant \cdot \delta}\) multiplied onto the maximum coefficient of a least squares solution to obtain the value if using the ‘reciprocal’ method (the default is 2.75).
- sigma_stop_fixed : float
- The minimum value of std. dev. in Gaussian l0 approx (the default 0.01).