Fréchet Derivatives of Implicit Functionals in Control Theory

2022 ◽  
pp. 79-93
2013 ◽  
Vol 2013 ◽  
pp. 1-16 ◽  
Author(s):  
Kairi Kasemets ◽  
Jaan Janno

We deduce formulas for the Fréchet derivatives of cost functionals of several inverse problems for a parabolic integrodifferential equation in a weak formulation. The method consists in the application of an integrated convolutional form of the weak problem and all computations are implemented in regular Sobolev spaces.


Geophysics ◽  
1990 ◽  
Vol 55 (12) ◽  
pp. 1589-1595 ◽  
Author(s):  
D. E. Boerner ◽  
J. S. Holladay

Fréchet derivatives play dual roles in electromagnetic (EM) methods as averaging functions relating conductivity to EM fields and as sensitivity functions relating conductivity perturbations to changes in these fields. For one‐dimensional EM inductive sounding, the Fréchet derivatives are not strongly model‐dependent, even for quite diverse earth models. In fact, using a scaled version of the Fréchet derivative for a uniform half‐space to approximate the exact Jacobian in a layered earth inversion program can actually improve the convergence to an acceptable model. This lack of a strong model dependence makes it possible to consider the capabilities and limitations of EM “imaging” methods from the perspective of Fréchet derivatives. Of particular interest is that noninductive Fréchet derivatives are strongly model‐dependent and hence the EM fields generated by this mode are less amenable to imaging techniques.


Geophysics ◽  
2001 ◽  
Vol 66 (5) ◽  
pp. 1364-1371
Author(s):  
Tsili Wang ◽  
Alberto Mezzatesta

Frechet derivatives provide the vital information for parametric resistivity inversion, but the calculation for a multidimensional problem is often computer intensive. This paper presents a new technique for fast calculation of the Frechet derivatives of resistivity measurements with respect to formation resistivity properties. The technique, referred to as the auxiliary source array method (ASAM), generalizes the reciprocity principle‐based methods in that for closely spaced receivers it may not be necessary to place a fictitious source at each receiver location. Rather, an auxiliary source array comprised of sparsely spaced fictitious sources can be constructed from which the field for any fictitious source location can be reconstructed. The ASAM was tested with a deviated‐well resistivity model for an array resistivity device that acquires 8 potential, 16 first potential difference, and 14 second potential difference data points at each depth level. The Frechet derivatives calculated by the ASAM agree well with those obtained through the parameter perturbation method. The tests showed that the calculation time of the ASAM has little dependence on the number of parameters for which the Frechet derivatives are to be calculated. The method can calculate the Frechet derivatives of 5 to 138 resistivity parameters with only 20% to 50% additional computer time. For the 138‐parameter model, the ASAM is about two orders of magnitude faster than the parameter perturbation method.


Sign in / Sign up

Export Citation Format

Share Document