Distributionally robust minimization in meta-learning for system identification
This work addresses safety-critical applications in system identification by improving robustness to worst-case scenarios, though it is incremental as it builds on existing meta-learning methods.
The paper tackled the problem of task variability in meta-learning for system identification by adopting a distributionally robust optimization paradigm that prioritizes high-loss tasks, resulting in enhanced performance in worst-case scenarios and reduced failures in safety-critical applications.
Meta learning aims at learning how to solve tasks, and thus it allows to estimate models that can be quickly adapted to new scenarios. This work explores distributionally robust minimization in meta learning for system identification. Standard meta learning approaches optimize the expected loss, overlooking task variability. We use an alternative approach, adopting a distributionally robust optimization paradigm that prioritizes high-loss tasks, enhancing performance in worst-case scenarios. Evaluated on a meta model trained on a class of synthetic dynamical systems and tested in both in-distribution and out-of-distribution settings, the proposed approach allows to reduce failures in safety-critical applications.