Towards a Bio-Inspired Real-Time Neuromorphic Cerebellum

Abstract

This work presents the first simulation of a large-scale, bio-physically constrained cerebellum model performed on neuromorphic hardware. A model containing 97,000 neurons and 4.2 million synapses is simulated on the SpiNNaker neuromorphic system. Results are validated against a baseline simulation of the same model executed with NEST, a popular spiking neural network simulator using generic computational resources and double precision floating point arithmetic. Individual cell and network-level spiking activity is validated in terms of average spike rates, relative lead or lag of spike times, and membrane potential dynamics of individual neurons, and SpiNNaker is shown to produce results in agreement with NEST. Once validated, the model is used to investigate how to accelerate the simulation speed of the network on the SpiNNaker system, with the future goal of creating a real-time neuromorphic cerebellum. Through detailed communication profiling, peak network activity is identified as one of the main challenges for simulation speed-up. Propagation of spiking activity through the network is measured, and will inform the future development of accelerated execution strategies for cerebellum models on neuromorphic hardware. The large ratio of granule cells to other cell types in the model results in high levels of activity converging onto few cells, with those cells having relatively larger time costs associated with the processing of communication. Organizing cells on SpiNNaker in accordance with their spatial position is shown to reduce the peak communication load by 41%. It is hoped that these insights, together with alternative parallelization strategies, will pave the way for real-time execution of large-scale, bio-physically constrained cerebellum models on SpiNNaker. This in turn will enable exploration of cerebellum-inspired controllers for neurorobotic applications, and execution of extended duration simulations over timescales that would currently be prohibitive using conventional computational platforms.

The quest for multiscale brain modeling

“…Various methods are available that allow multicompartment models to be compressed into single point neuron models while maintaining salient aspects of spike timing and dendritic computation [72][73][74][75]. SNNs composed of point neurons [76][77][78] have been used to simulate large-scale networks [70,71] and extend to closed-loop controllers [79,80] and neuromorphic computers [81]. SNNs can be further compressed by formalizing a transfer function that summarizes the statistical properties of the input-output relationship into mean-field (MF) models [82][83][84] that can be effective representations of the mesoscopic level.…”

Brain-Inspired Spiking Neural Network Controller for a Neurorobotic Whisker System

“…An already available solution to gain real-time performances can be to rely on spiking neural networks running on neuromorphic hardware. Very recently, a cerebellar-inspired model made of 97,000 neurons and 4.2 million synapses has been implemented on the neuromorphic platform SpiNNaker (Bogdan et al, 2021 ). This solution could be applicable if the plasticity rule used at PF-PC synapses, supervised by IO activity, will be implemented on this or other neuromorphic systems.…”

Brain-inspired spiking neural network controller for a neurorobotic whisker system

“…An already available so- lution to achieve real-time performances can be to rely on spiking neural networks running on neuromorphic hardware. Very recently, a cerebellar-inspired model made of 97 thousand neurons and 4.2 million synapses has been implemented on the neuromorphic platform SpiNNaker (65). This type of solution could be applicable if the plasticity rule used at PF-PC synapses, supervised by IO activity, will be implemented on this or other neuromorphic systems.…”