General Uncertainty Estimation with Delta Variances
This addresses uncertainty estimation for decision-makers in data-limited scenarios, offering an incremental improvement by unifying and extending existing methods.
The paper tackles the problem of efficiently estimating epistemic uncertainty in large neural networks, proposing Delta Variances as a family of algorithms that achieve competitive results with low computational cost, such as requiring only a single gradient computation in a weather simulator example.
Decision makers may suffer from uncertainty induced by limited data. This may be mitigated by accounting for epistemic uncertainty, which is however challenging to estimate efficiently for large neural networks. To this extent we investigate Delta Variances, a family of algorithms for epistemic uncertainty quantification, that is computationally efficient and convenient to implement. It can be applied to neural networks and more general functions composed of neural networks. As an example we consider a weather simulator with a neural-network-based step function inside -- here Delta Variances empirically obtain competitive results at the cost of a single gradient computation. The approach is convenient as it requires no changes to the neural network architecture or training procedure. We discuss multiple ways to derive Delta Variances theoretically noting that special cases recover popular techniques and present a unified perspective on multiple related methods. Finally we observe that this general perspective gives rise to a natural extension and empirically show its benefit.