This is a repository for the GEPNet MIMO detector https://arxiv.org/pdf/2201.03731.pdf

Guideline

Training

1. Set all training parameters in the parsers.py.
2. Start GEPNet training by running main_with_genData.py.
3. All training reports are saved in the directory reports --> GEPNet_"Nr"X"Nt_list"_ "num_constell"QAM. There will be four files saved during the training:
4. log.txt --> record elapsed time in each epoch.
5. current_acc.txt --> Record current validation SERs.
6. reportHPC.txt --> record number of epoch, loss, and training accuracy.
7. best_epoch.txt --> record the best epoch in terms of validation SER.
8. The trained model is saved in the directory models --> GEPNet_"Nr"X"Nt_list"_ "num_constell"QAM. There will be only one model that corresponds to the best epoch.

Testing

1. Set all testing parameters in the parsers.py.
2. Start GEPNet testing by running mainTestScript.py.
3. All testing reports are saved in the directory Test_reports --> GEPNet_"Nr"X"Nt_list"_ "num_constell"QAM. There will be two files saved during the testing:
4. NumberofTestingData.txt --> record number of testing data for each SNR point.
5. TestReport.txt --> record testing SNR and SER.
6. The testing is based on the trained model saved in the directory models --> GEPNet_"Nr"X"Nt_list"_"num_constell"QAM. There will be only one model that corresponds to the best epoch.