EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

GIANT Networks: Very Deep Fully Connected Neural Networks Applied to Microwave Problems

Simon Stenmark, Thomas Rylander, Tomas McKelvey, Andrei Ludvig‐Osipov

IET Microwaves, Antennas & Propagation · 2026-01 · 20 (1)

DOI: 10.1049/mia2.70077 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IET Microwaves, Antennas & Propagation, online erschienen am 07.03.2026. Verfasst von Simon Stenmark und 3 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: VINNOVA.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

We present the Gradient‐Informed Attentive Normalisation Training (GIANT) framework with the objective to create very deep fully connected neural networks, which we use as surrogate models in the context of microwave problems. As the central component of the GIANT framework, we introduce a novel dynamic reparameterisation procedure for the weight‐bias parameter space by means of a low‐variance preserving normalisation layer for each fully connected layer and we refer to this construction as Attentive Normalisation (AttNorm). As part of AttNorm, we also introduce a new and tailored updating scheme that improves the convergence during training. To efficiently train very deep fully connected neural networks, we exploit Sobolev training with gradient information, which is computed at a very low computational cost by means of continuum sensitivity analysis. We test our novel approach on two microwave applications: (i) a six‐port microwave cavity with a random medium and (ii) an H‐plane waveguide filter optimised under geometrical uncertainty. For these examples, we demonstrate successful training of neural networks with up to 30 layers, which are sufficiently accurate and expressive to serve as excellent surrogate models.

Förderung

  • VINNOVA

Zitieren

APA

Stenmark, S., Rylander, T., McKelvey, T., & Ludvig‐Osipov, A. (2026). GIANT Networks: Very Deep Fully Connected Neural Networks Applied to Microwave Problems. IET Microwaves, Antennas & Propagation, 20(1). https://doi.org/10.1049/mia2.70077

Vancouver

Stenmark S, Rylander T, McKelvey T, Ludvig‐Osipov A. GIANT Networks: Very Deep Fully Connected Neural Networks Applied to Microwave Problems. IET Microwaves, Antennas & Propagation. 2026;20(1). doi:10.1049/mia2.70077

BibTeX
@article{Stenmark2026,
  author = {Stenmark, Simon and Rylander, Thomas and McKelvey, Tomas and Ludvig‐Osipov, Andrei},
  title = {GIANT Networks: Very Deep Fully Connected Neural Networks Applied to Microwave Problems},
  journal = {IET Microwaves, Antennas & Propagation},
  year = {2026},
  volume = {20},
  number = {1},
  doi = {10.1049/mia2.70077}
}
Kostenlos starten →