"Uncertainty Quantification in Neural Differential Equations"
Graf, OlgaUncertainty quantification (UQ) helps to make trustworthy predictions based on collected observations and uncertain domain knowledge. With increased usage of deep learning in various applications, the need for efficient UQ methods that can make deep models more reliable has increased as well. Among applications that can benefit from effective handling of uncertainty are the deep learning based differential equation (DE) solvers. We adapt state-of-the-art UQ methods to get the predictive uncertainty for DE solutions by exploiting the link between network residuals, which are known and used as a loss function, and the absolute error of DE solution.