Memristor-Based Meta-Learning for Fast mmWave Beam Prediction in Non-Stationary Environments

TL;DR

Problem: mmWave beam prediction in non-stationary environments requires fast adaptation with limited data and multiple users ($N$ antennas, $K$ users) under a total transmit power $P$. Approach: a memristor-based meta-learning framework (M-ML) that provides a meta-initialization and uses a memory set of informative samples to enable rapid adaptation, coupled with a three-stage beam prediction pipeline and low-dimensional beamforming component decomposition. Contributions: memory-enabled data selection to mitigate overfitting and catastrophic forgetting, decomposition of the beamforming matrix into low-dimensional components $(w,u,\mu)$ with real-time reconstruction, and empirical demonstration of improved generalization in episodically dynamic channels. Findings: simulations show higher WSR and faster adaptation in unseen fading distributions (Rayleigh, Rician, Nakagami-$m$) than traditional MAML, unsupervised, or WMMSE baselines, with reduced data requirements and robust performance.

Abstract

Traditional machine learning techniques have achieved great success in improving data-rate performance and reducing latency in millimeter wave (mmWave) communications. However, these methods still face two key challenges: (i) their reliance on large-scale paired data for model training and tuning which limits performance gains and makes beam predictions outdated, especially in multi-user mmWave systems with large antenna arrays, and (ii) meta-learning (ML)-based beamforming solutions are prone to overfitting when trained on a limited number of tasks. To address these issues, we propose a memristorbased meta-learning (M-ML) framework for predicting mmWave beam in real time. The M-ML framework generates optimal initialization parameters during the training phase, providing a strong starting point for adapting to unknown environments during the testing phase. By leveraging memory to store key data, M-ML ensures the predicted beamforming vectors are wellsuited to episodically dynamic channel distributions, even when testing and training environments do not align. Simulation results show that our approach delivers high prediction accuracy in new environments, without relying on large datasets. Moreover, MML enhances the model's generalization ability and adaptability.

Figures (7)

• Figure 1: Schematic diagram for the system model.
• Figure 2: An overview of model-agnostic meta-learning framework.
• Figure 3: An overview of the training and testing phases of the M-ML framework.
• Figure 4: WSR performance comparison of distinct beam prediction approaches in Rician fading environment.
• Figure 5: WSR performance comparison of distinct beam prediction approaches in Rayleigh fading environment.