This paper examines different tomographic inversion techniques: 1) constrained tomography based on incorporated a priori information in the form of deterministic "hard' bounds; 2) tomographic modelling with both deterministic and stochastic (Gaussian) "soft' constraints; and in 3) the "hard' deterministic and "soft' Gibbs constraints are implemented simultaneously. In the case of Gaussian statistics, the stochastic constraints are fully determined by the covariance operators. Since the Gaussian statistics correspond to maximum entropy and thus can emerge as inefficient compared to other statistics, the possibility of incorporating stochastic "soft' constraints which do not follow the Gaussian distributions is explored. In particular, the Gibbs "soft' constraints are tested along with the deterministic "hard' constraints, and noticeable improvements in tomographic modelling are reported. -from Authors
Carrion, P., Jacovitti, G., Neri, A. (1993). Gaussian and non-Gaussian tomographic modelling via simulated annealing. JOURNAL OF SEISMIC EXPLORATION, 2(2), 189-204.
Gaussian and non-Gaussian tomographic modelling via simulated annealing
NERI, Alessandro
1993-01-01
Abstract
This paper examines different tomographic inversion techniques: 1) constrained tomography based on incorporated a priori information in the form of deterministic "hard' bounds; 2) tomographic modelling with both deterministic and stochastic (Gaussian) "soft' constraints; and in 3) the "hard' deterministic and "soft' Gibbs constraints are implemented simultaneously. In the case of Gaussian statistics, the stochastic constraints are fully determined by the covariance operators. Since the Gaussian statistics correspond to maximum entropy and thus can emerge as inefficient compared to other statistics, the possibility of incorporating stochastic "soft' constraints which do not follow the Gaussian distributions is explored. In particular, the Gibbs "soft' constraints are tested along with the deterministic "hard' constraints, and noticeable improvements in tomographic modelling are reported. -from AuthorsI documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.