Robust Hybrid Learning With Expert Augmentation
This work addresses the generalization issue in hybrid modelling for dynamical systems, offering a method to enhance robustness, though it appears incremental as it builds on existing hybrid systems.
The paper tackles the problem of hybrid models' limited performance guarantees outside the training distribution by introducing expert augmentation, a data augmentation strategy that leverages expert model validity to improve generalization, with empirical validation on controlled experiments and a real-world double pendulum dataset.
Hybrid modelling reduces the misspecification of expert models by combining them with machine learning (ML) components learned from data. Similarly to many ML algorithms, hybrid model performance guarantees are limited to the training distribution. Leveraging the insight that the expert model is usually valid even outside the training domain, we overcome this limitation by introducing a hybrid data augmentation strategy termed \textit{expert augmentation}. Based on a probabilistic formalization of hybrid modelling, we demonstrate that expert augmentation, which can be incorporated into existing hybrid systems, improves generalization. We empirically validate the expert augmentation on three controlled experiments modelling dynamical systems with ordinary and partial differential equations. Finally, we assess the potential real-world applicability of expert augmentation on a dataset of a real double pendulum.