Novel approaches to entropic regularization (Q4719142)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 1383373
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | Novel approaches to entropic regularization |
scientific article; zbMATH DE number 1383373 |
Statements
Novel approaches to entropic regularization (English)
0 references
1 February 2000
0 references
numerical examples
0 references
inverse problem
0 references
isotropic smoothing
0 references
entropic regularization
0 references
maximum entropy regularization method
0 references
geoelectric conductivity distributions
0 references
magnetotelluric data
0 references
atmosphere
0 references
radiosonde measurements
0 references
0.9044055
0 references
0.90087515
0 references
0.89575565
0 references
0.89423054
0 references
0.88300997
0 references
0 references
0.8719392
0 references
0.86847144
0 references
Many times, when solving an inverse problem, imposing an isotropic smoothing directly on the unknown parameters may lead to an unnecessary loss of resolution or to an unacceptable bias. In this work, two new entropic regularization techniques are introduced. They represent a generalization of the well-known maximum entropy regularization method, and allow of a greater flexibility for introducing any prior information about the expected structure of the true physical model, or its derivatives, into the inversion procedure. The first technique, denoted MinEnt-1, on the minimization of the entropy of the vector of first differences of the unknown parameters. Unlike other classical regularization schemes, which maximize smoothness for a given data, the MinEnt-1 method constrains the class of possible solutions into a restricted set of low entropy models, constituted by locally smooth regions separated by sharp discontinuities. Numerical simulations have shown that, when the true model has low entropy content, the MinEnt-1 algorithm converges to excellent reconstructions, yielding results that are superior to those obtained by the maximum entropy formalism. To illustrate the essential feature of the method, MinEnt-1 is applied to the reconstruction of two-dimensional geoelectric conductivity distributions from magnetotelluric data. The second technique is based on the maximization of the entropy of the vector of second differences of the unknown parameters and is denoted as the MaxEnt-2 method. For increasing values of the regularization parameter, MaxEnt-2 solutions tend to the second-order polynomial that better fits the data. Applied to the problem of retrieval of vertical temperature profiles in the atmosphere from remote sensing data, MaxEnt-2 results achieved good agreement with radiosonde measurements. The numerical simulations have also shown that MaxEnt-2 retrievals are relatively independent of the choice of the initial guess.
0 references