Neural computing with coherent laser networks

Abstract We show that coherent laser networks (CLNs) exhibit emergent neural computing capabilities. The proposed scheme is built on harnessing the collective behavior of for storing a number phase patterns as stable fixed points governing dynamical equations and retrieving such through proper excitation conditions, thus exhibiting an associative memory property. It discussed despite large storage capacity network, overlap between fixed-point effectively limits pattern retrieval to only two images. Next, we this restriction can be uplifted by using nonreciprocal coupling lasers allows utilizing capacity. This work opens new possibilities computation with novel analog processors. In addition, underlying model here suggests energy-based recurrent network handles continuous data opposed Hopfield Boltzmann machines are intrinsically binary systems.