Self-Supervised Pretraining for Transmission-Relevant MIMO CSI Representation
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 H, R=HHH determines the received power of any unit-norm transmit beam w as |Hw|22=wHRw. 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.
Related
- Self-Supervised Pretraining of Molecular Graph Encoders with LeJEPA
- Dual-domain self-supervised feature alignment via spectral–spatial representation learning for deepfake anomaly detection
- From Pixels, Without Pre-training: Joint Generative and Self-Supervised Representation Learning in One Model
- PhysBridge: Response-supervised grid pretraining with physics-constrained local correction for transferable power-system intelligence
- PermuFormer: Multi-Task Pretraining for Permutation Representation in Algebraic Combinatorics
BibTeX
@misc{ryu2026selfsupervised,
title = {{Self-Supervised Pretraining for Transmission-Relevant MIMO CSI Representation}},
author = {Sehyun Ryu and Yumin Kim and Minjae Lee and Hyun Jong Yang and John M. Cioffi},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2607.24872},
eprint = {2607.24872},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2607.24872},
}
RIS
TY - UNPB
TI - Self-Supervised Pretraining for Transmission-Relevant MIMO CSI Representation
AU - Sehyun Ryu
AU - Yumin Kim
AU - Minjae Lee
AU - Hyun Jong Yang
AU - John M. Cioffi
PY - 2026
DA - 2026/09/06
PB - arXiv
DO - 10.48550/arxiv.2607.24872
UR - https://arxiv.org/abs/2607.24872
AB - 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 H, R=H^HH determines the received power of any unit-norm transmit beam w as |Hw|_2^2=w^HRw. 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.
ER -