Schumann, Gerrit und Montag, Carsten und Steffens, Lars und Karl, Michael und Marx Gómez, Jorge (2025) Decentralized Federated Learning using Transformer-based Language Models. 2025 International Conference on Recent Advances in Information Systems (ICRAIS), 2025-09-10 - 2025-09-12, Mauritius.
![]() |
PDF
- Nur DLR-intern zugänglich
853kB |
Kurzfassung
When simulating federated learning scenarios, most studies use a centralized learning architecture in which a global server performs the orchestration of the individual clients and the aggregation of the associated weights. Decentralized federated learning (DFL) clusters, on the other hand, are rarely investigated, although they can counteract disadvantages such as "single point of failure" or limited scaling options. Furthermore, Transformer-based language models have not yet been investigated in DFL clusters, although they are still the de facto standard in the field of natural language processing. In order to investigate how such model architectures can be trained in a decentralized federated manner and which influencing factors play a decisive role, this study presents the implementation of three peer-to-peer topologies (Ring, Wheel, Fully connected) for a DFL cluster and their evaluation using the well-known text classification problem "Contradiction Detection". The results of this case study show that the use of the "Fully connected" topology leads to faster and more stable convergence (model divergence of 0.38), while "Ring" and "Wheel" topologies initially exhibit greater fluctuations (model divergence of up to 0.53).
elib-URL des Eintrags: | https://elib.dlr.de/215929/ | ||||||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Dokumentart: | Konferenzbeitrag (Vortrag) | ||||||||||||||||||||||||
Titel: | Decentralized Federated Learning using Transformer-based Language Models | ||||||||||||||||||||||||
Autoren: |
| ||||||||||||||||||||||||
Datum: | 2025 | ||||||||||||||||||||||||
Referierte Publikation: | Ja | ||||||||||||||||||||||||
Open Access: | Nein | ||||||||||||||||||||||||
Gold Open Access: | Nein | ||||||||||||||||||||||||
In SCOPUS: | Nein | ||||||||||||||||||||||||
In ISI Web of Science: | Nein | ||||||||||||||||||||||||
Status: | veröffentlicht | ||||||||||||||||||||||||
Stichwörter: | Decentralized Federated Learning, Peer-to-Peer, Gossip Protocol, Transformer Model, Text Classification, Contradiction Detection | ||||||||||||||||||||||||
Veranstaltungstitel: | 2025 International Conference on Recent Advances in Information Systems (ICRAIS) | ||||||||||||||||||||||||
Veranstaltungsort: | Mauritius | ||||||||||||||||||||||||
Veranstaltungsart: | internationale Konferenz | ||||||||||||||||||||||||
Veranstaltungsbeginn: | 10 September 2025 | ||||||||||||||||||||||||
Veranstaltungsende: | 12 September 2025 | ||||||||||||||||||||||||
HGF - Forschungsbereich: | Luftfahrt, Raumfahrt und Verkehr | ||||||||||||||||||||||||
HGF - Programm: | Verkehr | ||||||||||||||||||||||||
HGF - Programmthema: | keine Zuordnung | ||||||||||||||||||||||||
DLR - Schwerpunkt: | Verkehr | ||||||||||||||||||||||||
DLR - Forschungsgebiet: | V - keine Zuordnung | ||||||||||||||||||||||||
DLR - Teilgebiet (Projekt, Vorhaben): | V - keine Zuordnung | ||||||||||||||||||||||||
Standort: | Rhein-Sieg-Kreis | ||||||||||||||||||||||||
Institute & Einrichtungen: | Institut für KI-Sicherheit | ||||||||||||||||||||||||
Hinterlegt von: | Steffens, Lars | ||||||||||||||||||||||||
Hinterlegt am: | 25 Sep 2025 09:20 | ||||||||||||||||||||||||
Letzte Änderung: | 25 Sep 2025 09:20 |
Nur für Mitarbeiter des Archivs: Kontrollseite des Eintrags