Beam-Response Contrastive Learning for Transmitter-Side MIMO CSI Representation

2026-07-26

AI Summary

This work addresses a key limitation in existing self-supervised pretraining methods for MIMO channel state information (CSI), which fail to explicitly model transmit-side channel similarity, thereby constraining representation transferability. To overcome this, the paper proposes Beam Response Contrastive Learning (BRCL), the first approach to incorporate transmit beam response characteristics into CSI self-supervised representation learning. BRCL constructs beam response profiles via Gram matrices at the transmitter and leverages the induced soft similarity as an unlabeled relational target, jointly optimizing CSI sample reconstruction and beam response consistency. Notably, the method requires neither task-specific labels nor manually crafted positive sample pairs. Evaluated across multiple MIMO datasets, BRCL significantly improves label efficiency and outperforms baseline approaches—including autoencoders and channel charting—on downstream tasks such as beam selection, user scheduling, and future beam prediction.

Abstract

Self-supervised representation learning from unlabeled channel state information (CSI) can reduce labeling and adaptation overhead in learning-based multiple-input multiple-output (MIMO) systems. Existing CSI pretraining methods typically use reconstruction objectives or contrastive pairs from generic augmentations, which do not explicitly reflect transmission-relevant channel similarity. This paper proposes beam-response contrastive learning (BRCL), a self-supervised CSI pretraining framework based on the transmit-side Gram matrix. For a channel matrix $\mathbf{H}$, $\mathbf{R}=\mathbf{H}^{\mathrm{H}}\mathbf{H}$ determines the received power of any unit-norm transmit beam $\mathbf{w}$ as $|\mathbf{H}\mathbf{w}|_2^2=\mathbf{w}^{\mathrm{H}}\mathbf{R}\mathbf{w}$. BRCL maps each CSI sample to a beam-response profile and uses the induced soft similarity as a label-free relational target for contrastive pretraining. Combined with reconstruction learning, BRCL enforces both sample-level CSI recovery and beam-response-level consistency, yielding transferable CSI representations without task-specific labels or manual positive pairs. Experiments on diverse MIMO channel datasets show that BRCL improves label efficiency and outperforms autoencoder- and channel-charting-based pretraining across beam selection, user selection, and future beam selection tasks.

Problem

• MIMO
• CSI representation
• self-supervised learning
• beam selection
• contrastive learning

Innovation

• beam-response contrastive learning
• self-supervised representation learning
• MIMO CSI
• Gram matrix
• label-efficient pretraining