Prediction uncertainty and optimal experimental design for learning dynamical systems

TL;DR

This paper introduces prediction deviation, a novel metric for quantifying uncertainty in dynamical system models, and demonstrates its use in designing experiments to reduce this uncertainty, with applications in biological systems.

Contribution

The paper presents prediction deviation as a new uncertainty metric and develops a method to optimize experimental design to minimize this uncertainty in dynamical systems.

Findings

Prediction deviation effectively quantifies model uncertainty.

The method guides optimal experiment selection to reduce uncertainty.

Theoretical bounds relate prediction deviation to true model trajectories.

Abstract

Dynamical systems are frequently used to model biological systems. When these models are fit to data it is necessary to ascertain the uncertainty in the model fit. Here we present prediction deviation, a new metric of uncertainty that determines the extent to which observed data have constrained the model's predictions. This is accomplished by solving an optimization problem that searches for a pair of models that each provide a good fit for the observed data, yet have maximally different predictions. We develop a method for estimating a priori the impact that additional experiments would have on the prediction deviation, allowing the experimenter to design a set of experiments that would most reduce uncertainty. We use prediction deviation to assess uncertainty in a model of interferon-alpha inhibition of viral infection, and to select a sequence of experiments that reduces this uncertainty. Finally we prove a theoretical result which shows that prediction deviation provides bounds on the trajectories of the underlying true model. These results show that prediction deviation is a meaningful metric of uncertainty that can be used for optimal experimental design.