FUZZY NEURAL NETWORKS FUZZY NEURAL NETWORKS FUZZY NEURAL NETWORKS FUZZY NEURAL NETWORKS
This work investigates several aspects of comparato r networks and shows that the min-max model is the ‘strongest’ model of computation which obeys the 0-1 principle.
Abstract
This paper comparator networks - a well-known model of parallel computation. This model is used extensi vely for keys arrangement tasks such as sorting and sele ction. This work investigates several aspects of comparato r networks. It starts with presenting handy tools for analysis of comparator networks in the form of conclusive sets - non-binary vectors that verify a specific functionality. The 0-1 principle introduced by Knut h states that a comparator network is a sorting netwo rk if and only if it sorts all binary inputs. Hence, it p oints out a certain binary conclusive set. We compare these two models by considering several 0-1 -like principles and show that the min-max model is the ‘strongest’ model of computation which obeys our principles. That is, if a function is computable in a model of computation in which any of these principles holds, a min-max network can compute this function.