Two-Timescale-DNN

Two-Timescale End-to-End Learning for Channel Acquisition and Hybrid Precoding

This repository contains the entire code for our work "Two-Timescale End-to-End Learning for Channel Acquisition and Hybrid Precoding", available at: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9610037 and has been accepted for publication in IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS (JSAC).

For any reproduce, further research or development, please kindly cite our JSAC Journal paper:

Q. Hu, Y. Cai, K. Kang, G. Yu, J. Hoydis, and Y. C. Eldar, "Two-Timescale End-to-End Learning for Channel Acquisition and Hybrid Precoding," IEEE J. Sel. Areas Commun., vol. 40, no. 1, pp. 163-181, Jan. 2022.

Requirements

The following versions have been tested: Python 3.6 + Pytorch 1.9.0. But newer versions should also be fine.

Training and Testing

Firstly, run "Train_singletime.py" and save the well-trained model and analog beamformers (set the path at "torch.save(state, path)", "torch.save(FRF_container, 'path')", "torch.save(WRF_container, 'path')");

Then, run "Train_twotime.py" and load the well-trained model and analog beamforming (set the path at "FRF = torch.load('path')", "WRF = torch.load('path')","load_data1 = torch.load(path)", "load_data2 = torch.load(path)").

The introduction of each file

complex_matrix.py: Some complex matrix operations;

Channel_gen.py: The function of generating channel samples;

Model_singletime.py: The model of long-term DNN;

Model_twotime.py: The model of short-term DNN;

Train_singletime.py: Train long-term DNN;

Train_twotime.py: Train short-term DNN.