Salvato in:
| Autore principale: | |
|---|---|
| Natura: | Recurso digital |
| Lingua: | |
| Pubblicazione: |
Zenodo
2026
|
| Soggetti: | |
| Accesso online: | https://doi.org/10.5281/zenodo.18945715 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
Sommario:
- Optimization algorithms are frequently deployed in dynamic environments where the underlying data distribution may shift over time. This distribution shift can significantly degrade the performance of models trained on historical data. This paper proposes an adaptive regularization framework to mitigate the impact of data distribution shift on optimization performance. The proposed approach dynamically adjusts the regularization strength based on real-time estimates of the distribution discrepancy between training and deployment data. We demonstrate the effectiveness of our method through simulations and provide theoretical justifications for its convergence properties.