Heteroscedastic Neural Networks for Path Loss Prediction with Link-Specific Uncertainty
Fuente:
arXiv
Gespeichert in:
| 1. Verfasser: | |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911292337422336 |
|---|---|
| author | Ethier, Jonathan |
| author_facet | Ethier, Jonathan |
| contents | Traditional and modern machine learning-based path loss models typically assume a constant prediction variance. We propose a neural network that jointly predicts the mean and link-specific variance by minimizing a Gaussian negative log-likelihood, enabling heteroscedastic uncertainty estimates. We compare shared, partially shared, and independent-parameter architectures using accuracy, calibration, and sharpness metrics on blind test sets from large public RF drive-test datasets. The shared-parameter architecture performs best, achieving an RMSE of 7.4 dB, 95.1 percent coverage for 95 percent prediction intervals, and a mean interval width of 29.6 dB. These uncertainty estimates further support link-specific coverage margins, improve RF planning and interference analyses, and provide effective self-diagnostics of model weaknesses. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_23243 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Heteroscedastic Neural Networks for Path Loss Prediction with Link-Specific Uncertainty Ethier, Jonathan Machine Learning Signal Processing Traditional and modern machine learning-based path loss models typically assume a constant prediction variance. We propose a neural network that jointly predicts the mean and link-specific variance by minimizing a Gaussian negative log-likelihood, enabling heteroscedastic uncertainty estimates. We compare shared, partially shared, and independent-parameter architectures using accuracy, calibration, and sharpness metrics on blind test sets from large public RF drive-test datasets. The shared-parameter architecture performs best, achieving an RMSE of 7.4 dB, 95.1 percent coverage for 95 percent prediction intervals, and a mean interval width of 29.6 dB. These uncertainty estimates further support link-specific coverage margins, improve RF planning and interference analyses, and provide effective self-diagnostics of model weaknesses. |
| title | Heteroscedastic Neural Networks for Path Loss Prediction with Link-Specific Uncertainty |
| topic | Machine Learning Signal Processing |
| url | https://arxiv.org/abs/2511.23243 |