A Normalizing aVLSI Network with Controllable Winner-Take-All Properties

We describe an aVLSI network consisting of a group of excitatory neurons and a global inhibitory neuron. The output of the inhibitory neuron is normalized with respect to the input strengths in a manner that is useful in any system where we wish the output signal to code only the strength of the inputs, and not be dependent on the number of active inputs. The circuitry in each neuron is equivalent to that in Lazzaro’s winner-take-all (WTA) circuit [1] with one additional transistor and a voltage reference. As in Lazzaro’s circuit, the outputs of the excitatory neurons code for the neuron with the largest input. The novel feature is that multiple winners can be chosen (soft-max). By varying one parameter, the network can operate in a soft-max regime or a WTA regime. We show results from two different fabricated networks.