Spiking Neural Network Algorithms for Superconducting Optoelectronic Hardware

Neuromorphic computing seeks to draw inspiration from the brain in order to reinvent artificially intelligent systems with hopes to exploit the flexibility, speed, and energy efficiency of organic intelligence. This effort is defined by two major (and often disconnected) fronts—hardware development and algorithms. The prior defines the physical structure of the computational substrate in question and the latter determines the paradigm through which information is processed. Progress on both fronts is abundant, but less often is it synchronized. We here explore a deliberate approach of designing custom spiking neural network (SNN) algorithms on simulated superconducting optoelectronic network (SOEN) hardware. Our results show promise for this coordinated approach, with progress on complex computational tasks such as speech and image recognition. Moreover, this study may be leveraged to justify the physical development of SOENs and therefore bring to fruition their light-speed information processing and brain-level scalability.