Flexible Data Fusion (and Fission)

Abstract

An approach is described for developing methods for fusion: given how events A & B occurring by themselves influence some measure, estimate the influence (on that measure) of A and B occurring together. An example is combine the effects of evidence on the belief (likelihood) of some hypothesis. This approach also deals with the opposite problem of estimating the effects on a measure of A and B by themselves when only their combined effects are known: data fusion. The methods developed will both 1) try to make intuitive estimations at information not given, and 2) not conflict with any information given (unless it is inconsistent).