Feasibility study of real-time spiking neural network simulations on a swarm intelligence based digital architecture

Nature has proved to be a source of inspiration for engineering solutions. Spiking Neural Networks are exemplary from this perspective, due to the possibility to exploit them not only to simulate the biological networks of neurons but also to effectively work as classifiers and artificial intelligence systems. Another interesting nature-inspired paradigm is Swarm Intelligence, mainly applied to optimization problems and robotics, but also used to create digital architectures for array processing, with self-organization and fault-tolerance features. The aim of this paper is to evaluate if nature-inspired Swarm Intelligence based architectures can be effectively used to simulate biological neuronal assemblies through spiking neural networks models. A preliminary assessment of the proposed approach on low-end FPGA devices reveals near real-time capabilities with fully-connected neural networks composed of 64 Izhikevich neurons. A prospective analysis on larger platforms reveals the scalability of the proposed approach.