Learning to recognize visual stimuli in neuromorphic VLSI

We demonstrate learning in a neuromorphic recurrent attractor network distributed onto two VLSI chips. On a monitor we present some stimuli which are input to the network through the neuromorphic retina. Stimulation induces modification in the synaptic weights up to the point in which the selective reverberant states of activity are supported in the absence of stimulus. The network activity and the evolution of the synaptic matrix are monitored during learning. The visitor can draw his/her own stimuli, he/she can modify the learning parameters to teach the network to recognize them.