On the effect of scalability and normalization in physics-informed neural networks

Show simple item record
dc.contributor.authorKovář, P.
dc.contributor.authorFürst, J.
dc.contributor.editorRendl, Jan
dc.date.accessioned2026-04-30T09:15:50Z
dc.date.available2026-04-30T09:15:50Z
dc.date.issued2026
dc.description.abstract-translatedPhysics-informed neural networks (PINNs) have emerged as a powerful framework for solving problems governed by partial differential equations. However, their performance and training stability are often sensitive to problem scaling. In this work, the effect of scalability in physics-informed learning, with particular emphasis on the normalization of input variables is presented. Appropriate input normalization plays a critical role in balancing data-driven and physics-based loss terms, improving gradient propagation, and enhancing convergence behavior, that is shown through a series of numerical experiments. These findings highlight normalization as a key yet often overlooked component in the effective deployment of PINNs.en
dc.description.sponsorshipCZ.02.1.01/0.0/0.0/16019/0000826, SGS25/123/OHK2/3T/12en
dc.format6 s.cs
dc.format.mimetypeapplication/pdf
dc.identifier.isbn978-80-261-1352-2
dc.identifier.urihttp://hdl.handle.net/11025/67897
dc.language.isoenen
dc.publisherUniversity of West Bohemia in Pilsenen
dc.rights© University of West Bohemia in Pilsenen
dc.rights.accessopenAccessen
dc.subjectfyzikálně informované neuronové sítěcs
dc.subjectškálycs
dc.subjectnormalizacecs
dc.subject.translatedphysics-informed neural networksen
dc.subject.translatedscalesen
dc.subject.translatednormalizationen
dc.titleOn the effect of scalability and normalization in physics-informed neural networksen
dc.typekonferenční příspěvekcs
dc.typeconferenceObjecten
dc.type.versionpublishedVersionen
local.files.count2*
local.files.size4597699*
local.has.filesyes*

Files

Original bundle
Showing 1 - 2 out of 2 results
No Thumbnail Available
Name:
AM2026_proceedings-88-93.pdf
Size:
3.8 MB
Format:
Adobe Portable Document Format
Download
No Thumbnail Available
Name:
AM2026_proceedings-uvod.pdf
Size:
603.56 KB
Format:
Adobe Portable Document Format
Download
License bundle
Showing 1 - 1 out of 1 results
No Thumbnail Available
Name:
license.txt
Size:
1.71 KB
Format:
Item-specific license agreed upon to submission
Description:
Download

Collections

Proceedings of the 27th International Conference Applied Mechanics 2026
Proceedings of the 27th International Conference Applied Mechanics 2026