Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2512.10056 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915667969572864 |
|---|---|
| author | Namazi, Alireza Fathkouhi, Amirreza Dolatpour Shakeri, Heman |
| author_facet | Namazi, Alireza Fathkouhi, Amirreza Dolatpour Shakeri, Heman |
| contents | Autoregressive forecasting is central to predictive control in diabetes and hemodynamic management, where different operating zones carry different clinical risks. Standard models trained with teacher forcing suffer from exposure bias, yielding unstable multi-step forecasts for closed-loop use. We introduce Soft-Token Trajectory Forecasting (SoTra), which propagates continuous probability distributions (``soft tokens'') to mitigate exposure bias and learn calibrated, uncertainty-aware trajectories. A risk-aware decoding module then minimizes expected clinical harm. In glucose forecasting, SoTra reduces average zone-based risk by 18\%; in blood-pressure forecasting, it lowers effective clinical risk by approximately 15\%. These improvements support its use in safety-critical predictive control. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_10056 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Mitigating Exposure Bias in Risk-Aware Time Series Forecasting with Soft Tokens Namazi, Alireza Fathkouhi, Amirreza Dolatpour Shakeri, Heman Machine Learning Autoregressive forecasting is central to predictive control in diabetes and hemodynamic management, where different operating zones carry different clinical risks. Standard models trained with teacher forcing suffer from exposure bias, yielding unstable multi-step forecasts for closed-loop use. We introduce Soft-Token Trajectory Forecasting (SoTra), which propagates continuous probability distributions (``soft tokens'') to mitigate exposure bias and learn calibrated, uncertainty-aware trajectories. A risk-aware decoding module then minimizes expected clinical harm. In glucose forecasting, SoTra reduces average zone-based risk by 18\%; in blood-pressure forecasting, it lowers effective clinical risk by approximately 15\%. These improvements support its use in safety-critical predictive control. |
| title | Mitigating Exposure Bias in Risk-Aware Time Series Forecasting with Soft Tokens |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2512.10056 |