Hybrid-Task Meta-Learning: A GNN Approach for Scalable and Transferable Bandwidth Allocation

TL;DR

This paper tackles scalable bandwidth allocation under diverse QoS demands and non-stationary channels by introducing a GNN-based policy trained with a hybrid-task meta-learning (HML) framework. The approach uses feature-engineered inputs and a two-type-node GNN to produce allocation decisions that scale with the number of users, while meta-training across multiple tasks enables rapid adaptation to unseen scenarios with few samples. Empirical results show near-optimal performance and substantial reductions in inference time (up to 200–2000×) compared with iterative optimization, and faster convergence (8–30 epochs) than MAML or transfer-learning baselines. The work demonstrates strong generalization across channel models, QoS requirements, and system parameters, with potential extensions to other network-optimization problems in next-generation wireless systems.

Abstract

In this paper, we develop a deep learning-based bandwidth allocation policy that is: 1) scalable with the number of users and 2) transferable to different communication scenarios, such as non-stationary wireless channels, different quality-of-service (QoS) requirements, and dynamically available resources. To support scalability, the bandwidth allocation policy is represented by a graph neural network (GNN), with which the number of training parameters does not change with the number of users. To enable the generalization of the GNN, we develop a hybrid-task meta-learning (HML) algorithm that trains the initial parameters of the GNN with different communication scenarios during meta-training. Next, during meta-testing, a few samples are used to fine-tune the GNN with unseen communication scenarios. Simulation results demonstrate that our HML approach can improve the initial performance by 8.79%, and sample efficiency by 73%, compared with existing benchmarks. After fine-tuning, our near-optimal GNN-based policy can achieve close to the same reward with much lower inference complexity compared to the optimal policy obtained using iterative optimization. Numerical results validate that our HML can reduce the computation time by approximately 200 to 2000 times than the optimal iterative algorithm.

Figures (11)

• Figure 1: GNN-based scalable bandwidth allocation.
• Figure 2: Tasksets of meta-learning algorithms, where different shapes represent different tasks.
• Figure 3: Average training losses with different numbers of users, where the secrecy rate in the long blocklength regime is considered, $r_{\tau}^{S,\mathcal{I}}=10$ Mbps, and $W_{\tau}^{S,\mathcal{I}} = 100$ MHz.
• Figure 4: Testing samples are selected from taskset $\mathcal{T}^\mathrm{F}$ and $\mathcal{T}^\mathrm{E}$ in Table. \ref{['table_Simulation_task_parameters']}.
• Figure 5: Meta-testing with unseen channel models.