Spectrum sharing in Device-to-Device (D2D) communications with power control and link scheduling is a challenging non-convex combinatorial optimization problem. The state-of-the-art model-based iterative algorithms such as FPLinQ produce optimum-achieving solutions, whilst deep learning-based approaches have been proposed recently to approximate FPLinQ with reduced computational complexity. However, due to the highly non-convex nature of the optimization problem, FPLinQ exhibits certain deficiencies in the highly interference-limited networks, as it may be trapped within certain local sub-optimal solutions that may be far from the global optimum. To address these issues, we propose to unfold FPLinQ, with certain parameters inside the iterative procedure adjusted by a graph reinforcement learning (GRL) method, and end up with a novel hybrid model/data-driven approach, termed UFPLinQ. Not only does UFPLinQ inherit the advantages of FPLinQ and GRL with respect to local optimality, explainability, scalability, and generalizability, but it also provides excellent solutions in interference-limited networks where FPLinQ fails. By numerical evaluations, UFPLinQ outperforms existing learning-based power control mechanisms, with substantially reduced training samples and iterations, and more interestingly remedies the potential deficiencies of FPLinQ in highly interference-limited networks.

Device-to-device (D2D) spectrum sharing in wireless communications is a challenging non-convex combinatorial optimization problem, involving entangled link scheduling and power control in a large-scale network. The state-of-the-art methods, either from a model-based or a data-driven perspective, exhibit certain limitations such as the critical need for channel state information (CSI) and/or a large number of (solved) instances (e.g., network layouts) as training samples. To advance this line of research, we propose a novel hybrid model/data-driven spectrum sharing mechanism with graph reinforcement learning for link scheduling (GRLinQ), injecting information theoretical insights into machine learning models, in such a way that link scheduling and power control can be solved in an intelligent manner. Through an extensive set of experiments, GRLinQ demonstrates superior performance to the existing model-based and data-driven link scheduling and/or power control methods, with a relaxed requirement for CSI, a substantially reduced number of unsolved instances as training samples, a possible distributed deployment, reduced online/offline computational complexity, and more remarkably excellent scalability and generalizability over different network scenarios and system configurations.

The advance of topological interference management (TIM) has been one of the driving forces of recent developments in network information theory. However, state-of-the-art coding schemes for TIM are usually handcrafted for specific families of network topologies, relying critically on experts’ domain knowledge and sophisticated treatments. The lack of systematic and automatic generation of solutions inevitably restricts their potential wider applications to wireless communication systems, due to the limited generalizability of coding schemes to wider network configurations. To address such an issue, this work makes the first attempt to advocate revisiting topological interference alignment (IA) from a novel learning-to-code perspective. Specifically, we recast the one-to-one and subspace IA conditions as vector assignment policies and propose a unifying learning-to-code on graphs (LCG) framework by leveraging graph neural networks (GNNs) for capturing topological structures and reinforcement learning (RL) for decision-making of IA beamforming vector assignment. Interestingly, the proposed LCG framework is capable of recovering known one-to-one scalar/vector IA solutions for a significantly wider range of network topologies, and more remarkably of discovering new subspace IA coding schemes for multiple-antenna cases that are challenging to be handcrafted. The extensive experiments demonstrate that the LCG framework is an effective way to automatically produce systematic coding solutions to the TIM instances with arbitrary network topologies, and at the same time, the underlying learning algorithm is efficient with respect to online inference time and possesses excellent generalizability and transferability for practical deployment.

Device-to-Device (D2D) link scheduling in wireless communications is a challenging non-convex combinatorial optimization problem. The state-of-the-art methods, either from a model-based or a data-driven perspective, exhibit certain limitations such as the critical need of Channel State Information (CSI) and a large number of instances or solved instances as training samples. To advance this line of research, we propose a novel hybrid model/data-driven approach with Graph Reinforcement Learning for Link Scheduling (GRLinQ), injecting information theoretical insights into machine leaning models. GRLinQ demonstrates superior performance to the existing model-based and data-driven link scheduling mechanisms, with a relaxed requirement of CSI, a smaller number of unsolved instances as training samples, a possible distributed deployment, and more remarkably an excellent generalization ability over different network scenarios and system configurations.

The state-of-the-art coding schemes for topological interference management (TIM) problems are usually handcrafted for specific families of network topologies, relying critically on experts' domain knowledge. This inevitably restricts the potential wider applications to wireless communication systems, due to the limited generalizability. This work makes the first attempt to advocate a novel intelligent coding approach to mimic topological interference alignment via local graph coloring algorithms, leveraging the new advances of graph neural networks (GNNs) and reinforcement learning (RL). The extensive experiments demonstrate the excellent generalizability and transferability of the proposed approach, where the parameterized GNNs trained by small size TIM instances are able to work well on new unseen network topologies with larger size.