Transformer-based Scalable Beamforming Optimization via Deep Residual Learning
This work addresses real-time beamforming optimization for dynamic communication environments, offering a scalable solution with improved efficiency, though it is incremental as it builds on existing learning-to-optimize paradigms.
The paper tackles downlink beamforming optimization in large-scale MU-MISO channels by developing an unsupervised deep learning framework based on Transformers and residual connections, achieving faster inference than iterative methods and closely approaching WMMSE performance at high SNRs.
We develop an unsupervised deep learning framework for downlink beamforming in large-scale MU-MISO channels. The model is trained offline, allowing real-time inference through lightweight feedforward computations in dynamic communication environments. Following the learning-to-optimize (L2O) paradigm, a multi-layer Transformer iteratively refines both channel and beamformer features via residual connections. To enhance training, three strategies are introduced: (i) curriculum learning (CL) to improve early-stage convergence and avoid local optima, (ii) semi-amortized learning to refine each Transformer block with a few gradient ascent steps, and (iii) sliding-window training to stabilize optimization by training only a subset of Transformer blocks at a time. Extensive simulations show that the proposed scheme outperforms existing baselines at low-to-medium SNRs and closely approaches WMMSE performance at high SNRs, while achieving substantially faster inference than iterative and online learning approaches.