Dynamics of a winner-take-all neural network

This paper describes a neural network with lateral inhibition, which exhibits dynamic winner-take-all (WTA) behavior. The equations of this network model a current input MOSFET WTA circuit, which motivates the discussion. A very general sufficient condition for the network to have a WTA equilibrium point is obtained and sufficient conditions for the network to converge to the WTA point are presented. This gives explicit expressions for the resolution and lower bound of the input currents. We also show that whenever the network gets into the WTA region, it will stay in that region and settle down exponentially fast to the WTA point. This provides a speed up procedure for the decision making: as soon as it gets into the region, the winner can be picked up. Finally, we show that this WTA neural network has a self-resetting property. Copyright (C) 1996 Elsevier Science Ltd