Multi-Task Learning Architectures for Joint Interference Detection and KPI Prediction in 5G Networks
Resumo
Real-time interference detection and accurate Key Performance Indicator (KPI) prediction are critical for optimizing 5G radio access networks. Jointly addressing these objectives offers a comprehensive view of network behavior but presents a significant challenge: simultaneously optimizing for heterogeneous tasks—discrete interference classification and continuous KPI regression—often leads to negative transfer or inefficient parameterization. In this paper, we systematically investigate the effectiveness of Multi-Task Learning (MTL) for this dual objective by evaluating distinct architectural strategies, including Hard Parameter Sharing, Cross-Stitch networks, Multi-gate Mixture-of-Experts (MMoE), Progressive Layered Extraction (PLE), and an attention-based model, against a Single-Task Learning (STL) baseline. Using a dataset collected from a realistic 5G testbed, we analyze the trade-offs between classification accuracy, regression error, model complexity, and inference latency. Our experimental results demonstrate that no single architecture dominates across all metrics. While STL achieves high predictive performance, it is computationally prohibitive for real-time applications due to redundant feature extraction. Conversely, Hard Parameter Sharing offers minimal latency but suffers severe performance degradation due to rigid representation sharing. PLE delivers the highest classification accuracy (87.62%) but at the cost of increased model size. Ultimately, MMoE emerges as the optimal architecture for practical deployment; it achieves the lowest total test loss and high classification accuracy (86.80%) while reducing Floating Point Operations (FLOPs) by approximately 74% compared to STL, making it well-suited for practical interference-aware monitoring and optimization in 5G radio access networks.
Referências
FCCLab (2025). Fcciq dataset 01 – detailed methodology. [link]. Accessed: 2025-12-14.
Goodfellow, I., Bengio, Y., Courville, A., and Bengio, Y. (2016). Deep learning, volume 1. MIT press Cambridge.
Jagannath, A. and Jagannath, J. (2022). Multi-task learning approach for modulation and wireless signal classification for 5g and beyond: Edge deployment via model compression. Physical Communication, 54:101793.
Ma, J., Zhao, Z., Yi, X., Chen, J., Hong, L., and Chi, E. H. (2018). Modeling task relationships in multi-task learning with multi-gate mixture-of-experts. In Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining, pages 1930–1939.
Misra, I., Shrivastava, A., Gupta, A., and Hebert, M. (2016). Cross-stitch networks for multi-task learning. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 3994–4003.
Parera, C., Redondi, A. E., Cesana, M., Liao, Q., and Malanchini, I. (2019). Transfer learning for channel quality prediction. In 2019 IEEE International Symposium on Measurements & Networking (M&N), pages 1–6. IEEE.
Tang, H., Liu, J., Zhao, M., and Gong, X. (2020). Progressive layered extraction (ple): A novel multi-task learning (mtl) model for personalized recommendations. In Proceedings of the 14th ACM conference on recommender systems, pages 269–278.
Tommy, I., Li, X., and Qian, L. (2025). Spatio-temporal beam-level traffic forecasting in 5g wireless systems using multi-task learning. In 2025 IEEE International Conference on Machine Learning for Communication and Networking (ICMLCN), pages 1–7. IEEE.
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., and Polosukhin, I. (2017). Attention is all you need. Advances in neural information processing systems, 30.
Wang, P., Zhang, X., Ma, Y., Zhu, H., Jiao, J., and Zhang, Q. (2024). Mtl-srn: Multi-task learning-based signal recognition network. In GLOBECOM 2024-2024 IEEE Global Communications Conference, pages 2918–2923. IEEE.
Zhang, L., Yang, Q., Liu, X., and Guan, H. (2022). Rethinking hard-parameter sharing in multi-domain learning. In 2022 IEEE International Conference on Multimedia and Expo (ICME), pages 01–06. IEEE.
