Model uncertainty carries with it a loss of performance when finding an optimal operator. The IBR principle is to find an operator that, based on a cost function, is optimal over an uncertainty class relative to a prior (or posterior) distribution reflecting the state of knowledge regarding the underlying physical processes. While an IBR operator is optimal over the uncertainty class, it is likely to be suboptimal relative to the full (true) model. This loss of performance is the cost of uncertainty. Experiments can reduce model uncertainty. Since experiments can be costly and time consuming, the question arises as to which experiment can best reduce the uncertainty as it pertains to the operational objective, not necessarily as it pertains to the model uncertainty in general, for instance, the entropy.
Online access to SPIE eBooks is limited to subscribing institutions.