Resolution Estimates for Selected Coordinate Descent: Identification of Seismic Structure in the Area of Geothermal Plants

Part 9: System Identification for Manufacturing Control Applications International audience The coordinate descent method is a traditional inverse solver to optimization problems. In modern sectors of production research: in computer graphics, computer tomography, a theory of pattern recognition var...

Full description

Bibliographic Details
Main Authors: Smaglichenko, Tatyana, Smaglichenko, Alexander
Other Authors: Research Oil and Gas Institute, Russian Academy of Science (RAS), Trapeznikov Institute of Control Sciences (ICS RAS), Russian Academy of Sciences Moscow (RAS), Alexandre Dolgui, Alain Bernard, David Lemoine, Gregor von Cieminski, David Romero, TC 5, WG 5.7
Format: Conference Object
Language:English
Published: HAL CCSD 2021
Subjects:
Coordinate descent;Linear equations;Tomography;Structure recognition
[INFO]Computer Science [cs]
Krafla
Iceland
Online Access:https://hal.inria.fr/hal-04030369
https://hal.inria.fr/hal-04030369/document
https://hal.inria.fr/hal-04030369/file/509923_1_En_62_Chapter.pdf
https://doi.org/10.1007/978-3-030-85874-2_62
Description
Summary:Part 9: System Identification for Manufacturing Control Applications International audience The coordinate descent method is a traditional inverse solver to optimization problems. In modern sectors of production research: in computer graphics, computer tomography, a theory of pattern recognition various algorithms of the coordinate descent have been applied. In this paper, we investigate the novel algorithm of selected coordinate descent and outline the difference between this algorithm and the classical coordinate descent. The solution selection is performed owing to the search of the maximum among values of the specific parameter. The maximum indicates a single direction, which is responsible for the minimum of the function in the least square sense. We develop the technique for defining the explicit expression for the resolution measure of the linear systems, which are solved using the proposed algorithm. The algorithm and its resolution tool are applied to seismic observations collected in the area of the Krafla and Theistareykir geothermal power plants, northern Iceland. The result confirms that the distinctive feature of the algorithm is its effectiveness when large-size structures are retrieved. The analysis of the resolution parameter values shows that the calculation of this parameter might be helpful to recognize the true structure.

Similar Items