Embedding Model-Based Fast Meta Learning for Downlink Beamforming Adaptation

Juping Zhang,Yi Yuan,Gan Zheng,Ioannis Krikidis,Kai-Kit Wong

doi:10.1109/twc.2021.3094162

Abstract

This paper studies the fast adaptive beamforming for the multiuser multiple-input single-output downlink. Existing deep learning-based approaches assume that training and testing channels follow the same distribution which causes task mismatch, when the testing environment changes. Although meta learning can deal with the task mismatch, it relies on labelled data and incurs high complexity in the pre-training and fine tuning stages. We propose a simple yet effective adaptive framework to solve the mismatch issue, which trains an embedding model as a transferable feature extractor, followed by fitting the support vector regression. Compared to the existing meta learning algorithm, our method does not necessarily need labelled data in the pre-training and does not need fine-tuning of the pre-trained model in the adaptation. The effectiveness of the proposed method is verified through two well-known applications, i.e., the signal to interference plus noise ratio balancing problem and the sum rate maximization problem. Furthermore, we extend our proposed method to online scenarios in non-stationary environments. Simulation results demonstrate the advantages of the proposed algorithm in terms of both performance and complexity. The proposed framework can also be applied to general radio resource management problems.

Highlights

Beamforming is one of the most promising multi-antenna techniques that can realize the antenna diversity gain and mitigate multiuser interference simultaneously
It is based on the intuitive idea that the mapping from channel state to beamforming can be learned by training a neural network model in an offline manner, and the beamforming solution can be directly predicted using the trained model in real time
To further investigate the effectiveness of the proposed fast learning framework on beamforming design in non-stationary scenarios, we extend our framework to the online application to solve the beamforming prediction problem in real-time communications systems

Summary

Introduction

Beamforming is one of the most promising multi-antenna techniques that can realize the antenna diversity gain and mitigate multiuser interference simultaneously. The deep learning technique has been proposed to address the complexity of beamforming design using the ‘learning to optimize’ framework [5] It is based on the intuitive idea that the mapping from channel state to beamforming can be learned by training a neural network model in an offline manner, and the beamforming solution can be directly predicted using the trained model in real time. An obvious solution is to re-train the model by using newly collected data from the new environment This is impractical because the changing network does not allow enough time to collect enough new data and train a new model before violating the latency constraint. It is a pressing research challenge in multiantenna communications to achieve fast adaptation of beamforming solutions

Objectives

Results

Conclusion