Reasoning With Co-Variations

Adaptation is what allows a system to maintain consistent behavior across variations in operating environments. In some previous work, a symbolic representation of the variations between two or more elements of a set was proposed. This article goes one step further and defines co-variations as functional dependencies between variations. This gives us a natural deduction rule on variations, which we show can be easily extended to perform similarity-based reasoning. A method is also proposed to learn co-variations from the data. In this method, co-variations correspond to object implication rules in a pattern structure.