Transformer-based Scalable Beamforming Optimization via Deep Residual Learning

TL;DR

An unsupervised deep learning framework for downlink beamforming in large-scale MU-MISO channels is developed, allowing real-time inference through lightweight feedforward computations in dynamic communication environments and achieving substantially faster inference than iterative and online learning approaches.

Abstract

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.

Figures (5)

• Figure 1: Semi-amortized Learning-to-Optimize Beamforming Optimization Scheme.
• Figure 2: Architecture of a single Transformer block.
• Figure 3: The behaviors of the proposed multi-layer Transformer beamforming scheme.
• Figure 4: Ablation studies of different training strategies.
• Figure 5: Sum rate versus SNR of different beamforming schemes.