Towards real-time spiking neural networks simulations

Together with SpikeFORCE partners, we have developed a brain-inspired software-hardware hybrid system for adaptive motor control. The system mimicked the response of neurons in the cerebellum by computing their impulses. The cerebellum is a brain area well known in particular for its importance in the learning and coordination of motor activities (movements). The hybrid system combined the responses of different types of neurons that were generated on a hardware chip and by computer software. The two communicated the occurrence of spikes through the computer bus interface.

Moreover, SpikeFORCE has constructed a program for simulating spiking neural networks that can achieve real-time simulation in some cases. It is a table-based event-triggered simulator, which makes use of predictive threshold computation to speed up simulation. We have used it to simulate an adaptive cerebellar model composed of thousands of neurons learning in real-time the smooth tracking of different target trajectories in multiple contexts.