Zur Hauptnavigation wechseln Zur Suche wechseln Zum Hauptinhalt wechseln

Domain-as-Particle with PSO Methods for Neural-Network Feature Weighting

  • Fabio Berberi*
  • , Paolo Mercorelli
  • *Korrespondierende/r Autor/-in für diese Arbeit

Publikation: Beiträge in Zeitschriften › Konferenzaufsätze in Fachzeitschriften › Forschung › Begutachtung

Abstract

We present a framework that integrates Particle Swarm Optimization (PSO), machine learning, K-Fold cross-validation, and surrogate modeling to identify optimal weight vectors for feature scaling in neural network training. In our approach, the n-dimensional weight space is partitioned into non-overlapping subdomains, each corresponding to a PSO particle. Particle movement is guided by a characteristic vector, which is determined by the best-performing candidates in each subdomain and by information exchanged with neighboring regions. To reduce evaluation costs, a surrogate model—trained on a uniformly sampled subset of candidates—pre-filters particles before full K-Fold validation. The top candidates then undergo comprehensive validation, updating the characteristic vectors for subsequent iterations. This domain-as-particle PSO framework enables efficient weight discovery, significantly reducing computational overhead while maintaining robust performance. The effectiveness of this approach is demonstrated on real-world datasets.

OriginalspracheEnglisch
ZeitschriftAnnals of Computer Science and Intelligence Systems
Jahrgang45
Seiten (von - bis)41-47
Seitenumfang7
DOIs
PublikationsstatusErschienen - 2025
Veranstaltung20th Conference on Computer Science and Intelligence Systems - FedCSIS 2025: Computer Science and Intelligence Systems - Kraków, Poland, Kraków, Polen
Dauer: 14.11.2025 → 17.11.2025
Konferenznummer: 20
https://annals-csis.org/Volume_44/ (Proceedings URL)
https://doi.org/10.15439/978-83-973291-8-8 (DOI)
https://annals-csis.org/Volume_45/ (Proceedings URL)
https://doi.org/10.15439/978-83-973291-9-5 (DOI)
https://annals-csis.org/Volume_43/ (Proceedings URL)
https://doi.org/10.15439/978-83-973291-6-4 (DOI)

Bibliographische Notiz

Publisher Copyright:
© PTI 2024.

Fachgebiete und Schlagwörter

  • Ingenieurwissenschaften

ASJC Scopus Sachgebiete

  • Artificial intelligence
  • Angewandte Informatik
  • Information systems
  • Informationssysteme und -management

Fingerprint

Untersuchen Sie die Forschungsthemen von „Domain-as-Particle with PSO Methods for Neural-Network Feature Weighting“. Zusammen bilden sie einen einzigartigen Fingerprint.

Dieses zitieren