Computation on a spiking neural network created by degenerate optical parametric oscillators

We have developed an optical spiking neural network system created by degenerated optical parametric oscillators (DOPOs). The DOPO spiking neuron is created from two anti-symmetrically coupled DOPO pulses and can reproduce two types of firing behavior, which are categorized as Hodgkins' class-I and class-II spiking modes. Adaptive changes in spiking frequencies are induced by the inter-neuron interactions combined with the controllability of the spiking modes. Our system can be effectively applied to Ising problem solvers, because adaptive changes in the spiking frequencies make convergence faster. This potential of the DOPO spiking neurons opens up new avenues for exploring optical computing systems.