This repository is the Python implementation of paper "Recursive Flow: A Generative Framework for MIMO Channel Estimation"
train_flow.py: The train code, used to train the flow backbone.
test_rcflow.py: The test function. Can be directly run to get the results.
loaders.py: This file contains the dataloader.
Paper.pdf: This file is the PDF file of the paper.
results: This directory cotains the pretrained models.
bin: This directory cotains the channel data.
The channel data can be obtained from this repository: "score-based-channels"
Should you find this work beneficial, kindly grant it a star!
To follow our research, please consider citing:
Jiang Z, Zhu F, Huang C, et al. Recursive Flow: A Generative Framework for MIMO Channel Estimation. arXiv preprint arXiv:2601.15767, 2026.
@article{jiang2026recursive,
title={Recursive Flow: A Generative Framework for MIMO Channel Estimation},
author={Jiang, Zehua and Zhu, Fenghao and Huang, Chongwen and Jin, Richeng and Yang, Zhaohui and Chen, Xiaoming and Zhang, Zhaoyang and Debbah, M{\'e}rouane},
journal={arXiv preprint arXiv:2601.15767},
year={2026}
}