Abstract
Abstract
A parallel sorting system based on discrete-time K-Winners-Take-All (KWTA) neural circuit (NC) model is designed. The system is described by a set of difference equations and by step functions. The KWTA NC model of discrete-time is given by difference equation and by step functions. Corresponding functional block-diagrams of the system are presented. The upper limit on the number of iterations required to achieve a convergence of the sorting process to the steady state is defined. The system does not need a knowledge of range of input data change. In order to use the system, the minimal difference between the inputs should be known. The system is suitable for processing unknown inputs of finite values located in arbitrary unknown finite range. The system is characterized by arbitrary finite resolution of inputs, high speed, moderate computational complexity and complexity of software implementation. The results of computer simulations illustrating efficiency of the system are provided.