Label Noise Resistant $f$-Divergence Receiver
for Power Line Communications

Nicola Novello and Andrea M. Tonello

Official repository of the paper "Label Noise Resistant $f$-Divergence Receiver for Power Line Communications" published at IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids (SmartGridComm) 2024.

Novel class of neural receivers based on the $f$-divergence, with analysis in the presence of label noise. The specific application considered is Power Line Communications.

---

📷 Scenario

---

💻 How to run the code

The folder with the scripts must comprise a folder named Dataset containing the .mat file with the channel measurements.

The file main.py runs the experiments:

python3 main.py --noisy True --noise_type symm --noise_rate 0.1 

where "noisy" must be set to True to run the tests in the presence of label noise. "noise_type" can be: "symm" for symmetric noise, "sparse" for sparse noise, and "unif" for uniform noise. "noise_rate" must be a float with suggested values: 0.1, 0.2.

The scripts main_functions.py and utils.py define the functions needed in main.py.

---

📝 References

If you use the code for your research, please cite our paper:

@inproceedings{novello2024label,
  title={Label Noise Resistant f-Divergence Receiver for Power Line Communications},
  author={Novello, Nicola and Tonello, Andrea M},
  booktitle={2024 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids (SmartGridComm)},
  pages={517--522},
  year={2024},
  organization={IEEE}
}

---

📋 Acknowledgments

The implementation is based on / inspired by:

• https://github.com/nicolaNovello/discriminative-classification-fDiv

---

📧 Contact

nicola.novello@aau.at