A discrete-time recurrent neural network with one neuron for k-winners-take-all operation

Abstract

In this paper, a discrete-time recurrent neural network with one neuron and global convergence is proposed for k-winners-take-all (kWTA) operation. Comparing with the existing kWTA networks, the proposed network has simpler structure with only one neuron. The global convergence of the network can be guaranteed for kWTA operation. Simulation results are provided to show that the outputs vector of the network is globally convergent to the solution of the kWTA operation. © 2009 Springer Berlin Heidelberg.