Meta Learner-Based Transfer Learning: Bridging Simulation and Actual Router Metrics
Future carrier networks for the sixth generation of mobile communications are expected to guarantee performance across heterogeneous networks that cover multiple technologies. This integration requires the effective verification of network performance under unpredictable traffic conditions. To addre...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10979298/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Future carrier networks for the sixth generation of mobile communications are expected to guarantee performance across heterogeneous networks that cover multiple technologies. This integration requires the effective verification of network performance under unpredictable traffic conditions. To address this challenge, a router modeling approach defined as the data-driven estimation of actual router performance has been proposed to facilitate router performance verification. However, the creation of router models has been limited by the high cost and difficulty of acquiring the extensive real-world datasets for various traffic patterns required to train actual router models. Our motivation is to enable the digital verification of router performance with limited real-world datasets. This paper proposes a meta learner-based transfer learning method to estimate actual router metrics using models trained on network simulation data. The proposed method aims to build a model of actual router metrics based on limited real-world datasets by supplementing them with different ranges and types of simulation data. The proposed method addresses the differences in these datasets by utilizing Neural Processes as a meta-learner combined with Partial Least Squares analysis to capture and bridge representations between simulation and real-world datasets for packet multiplexing tasks for transfer learning. The results show that the proposed method improves the estimated accuracy of actual router metrics: throughput, packet loss rate, and packet delay. Additionally, the proposed method demonstrates robustness and effectiveness in scenarios where real-world data is limited while maintaining lower computational complexity compared to conventional methods. Our results suggest that this approach may assist network operators in estimating network performance, even with limited real-world data. |
|---|---|
| ISSN: | 2169-3536 |