Progressive and constant-speed order filtering neural network

In this paper, a new order filtering neural network, which can select a specific ordered value from all inputs, is developed and analyzed. The proposed neural net in two-layer structure iteratively converges to the solution with low and constant convergent speed, which is independent of the number of inputs. With progressive behavior, the proposed neural net obtains the more accurate result when the number of iterations increases if the derived convergent condition is satisfied. From the view points of convergence speed and hardware complexity, the proposed order filtering neural network is suitable for various applications.