Parallel computing structures capable of flexible associations and recognition of fuzzy inputs

We experimentally show that computing with attractors leads to fast adaptive behavior in which dynamical associations can be made between different inputs which initially produce sharply distinct outputs. We do so by first defining a set of simple local procedures which allow a computing array to change its state in time so as to produce classical Pavlovian conditioning. We then examine the dynamics of coalescence and dissociation of attractors with a number of quantitative experiments. We also show how such arrays exhibit generalization and differentiation of inputs in their behavior.