CNN models of current mode neuromorphic networks

This paper describes the use of the cellular neural network model in understanding the operation of current mode neuromorphic networks consisting of subthreshold MOS transistors. The original proof of stability for symmetric cellular neural networks relied upon the definition of a suitable Lyapunov function. Here we show that Lyapunov functions can be defined for arbitrary networks of subthreshold MOS transistors operating as diffusers or pseudo-conductances and for the winner-take-all circuit of Lazzaro. We also derive an expression for the resolution of the winner-take-all circuit.