Improving the Flexibility of Shape-Constrained Symbolic Regression with Extended Constraints

David Piringer, Stefan Wagner, Christian Haider, Armin Fohler, Siegfried Silber, Michael Affenzeller

Publikation: Beitrag in Buch/Bericht/TagungsbandKonferenzbeitragBegutachtung


We describe an approach to utilize a broader spectrum of domain knowledge to model magnetization curves for high magnetic field strengths 0 ≤ H≤ 10 6 with access to data points far below the saturation polarization. Thereby, we extend the implementation of Shape-Constrained Symbolic Regression. The extension allows the modification of model estimates by a given expression to apply additional sets of constraints. We apply the given expression of an Extended Constraint row-by-row and compare the minimum and maximum outputs with the target interval. Furthermore, we introduce regions and thresholds as additional tools for constraint description and soft constraint evaluation. Our achieved results demonstrate the positive impact of such additional knowledge. The logical downside is the dependence on that knowledge to describe applicable constraints. Nevertheless, the approach is a promising way to reduce the human calculation effort for extrapolating magnetization curves. For future work, we plan to combine soft and hard constraint evaluation as well as the utilization of structure template GP.

TitelComputer Aided Systems Theory – EUROCAST 2022
Untertitel18th International Conference
Redakteure/-innenRoberto Moreno-Díaz, Franz Pichler, Alexis Quesada-Arencibia
Herausgeber (Verlag)Springer
ISBN (Print)9783031253119
PublikationsstatusVeröffentlicht - 2022
Veranstaltung18th International Conference on Computer Aided Systems Theory, EUROCAST 2022 - Las Palmas de Gran Canaria, Spanien
Dauer: 20 Feb. 202225 Feb. 2022


NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Band13789 LNCS
ISSN (Print)0302-9743
ISSN (elektronisch)1611-3349


Konferenz18th International Conference on Computer Aided Systems Theory, EUROCAST 2022
OrtLas Palmas de Gran Canaria


Untersuchen Sie die Forschungsthemen von „Improving the Flexibility of Shape-Constrained Symbolic Regression with Extended Constraints“. Zusammen bilden sie einen einzigartigen Fingerprint.