Deep Learning-Prompted TCP Congestion Control to Explore High-Error Link Transmission Potential
IEEE Access · 2026 · 14 · 50856-50877
DOI: 10.1109/access.2026.3672345 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Marcelo Reis Silva und 1 weiteren Autor:innen. Die Arbeit nennt 38 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Coordination for the Improvement of Higher Education Personnel (CAPES)—Brazil.
Abstract (Originalsprache)
The error rate on Transmission Control Protocol/Internet Protocol (TCP/IP) networks can suddenly increase, even over modern connections. In this context, this work proposes a CC mechanism to increase transmission data over high-bandwidth TCP networks, threatened by high ( $\gt 10^{-5}$ ) Bit Error Rate (BER). For this purpose, this paper examines the learning capacity of Deep Learning Neural Networks (DLNNs) in real, high-bandwidth, and heavily trafficked networks. In addition, the technical core of the proposed method integrates DLNN models with a kernel module to measure the gains that DLNN-derived CC can provide in concrete links with significant error rates. The results underscore the impressive potential of the new mechanism, highlighting classifiers with accuracy above 95%, and achieving throughput gains with peaks exceeding five times that of the CC traditionally embedded in most systems, which promises to enhance unforeseen high-error communication in the future.
Förderung
- Coordination for the Improvement of Higher Education Personnel (CAPES)—Brazil
Zitieren
APA
Silva, M. R., & Marcondes, C. A. C. (2026). Deep Learning-Prompted TCP Congestion Control to Explore High-Error Link Transmission Potential. IEEE Access, 14, 50856-50877. https://doi.org/10.1109/access.2026.3672345
Vancouver
Silva MR, Marcondes CAC. Deep Learning-Prompted TCP Congestion Control to Explore High-Error Link Transmission Potential. IEEE Access. 2026;14:50856-50877. doi:10.1109/access.2026.3672345
BibTeX
@article{Silva2026,
author = {Silva, Marcelo Reis and Marcondes, Cesar A. C.},
title = {Deep Learning-Prompted TCP Congestion Control to Explore High-Error Link Transmission Potential},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {50856--50877},
doi = {10.1109/access.2026.3672345}
}
EUVIMED