NeuroFlow: A General Purpose Spiking Neural Network Simulation Platform using Customizable Processors

Kit Cheung, Simon R. Schultz, Wayne Luk
2016 Frontiers in Neuroscience  
13 NeuroFlow is a scalable spiking neural network simulation platform for off-the-shelf high 14 performance computing systems using customizable hardware processors such as Field-15 Programmable Gate Arrays (FPGAs). Unlike multi-core processors and application-specific 16 integrated circuits, the processor architecture of NeuroFlow can be redesigned and reconfigured to 17 suit a particular simulation to deliver optimised performance, such as the degree of parallelism to 18 employ. The
more » ... n process supports using PyNN, a simulator-independent neural network 19 description language, to configure the processor. NeuroFlow supports a number of commonly used 20 current or conductance based neuronal models such as integrate-and-fire and Izhikevich models, and 21 the spike-timing-dependent plasticity (STDP) rule for learning. A 6-FPGA system can simulate a 22 network of up to approximately 600,000 neurons and can achieve a real-time performance of 400,000 23 neurons. Using one FPGA, NeuroFlow delivers a speedup of up to 33.6 times the speed of an 8-core 24 processor, or 2.83 times the speed of GPU-based platforms. With high flexibility and throughput, 25 NeuroFlow provides a viable environment for large-scale neural network simulation. 26 27
doi:10.3389/fnins.2015.00516 pmid:26834542 pmcid:PMC4712299 fatcat:r733lgxhgjcppcw7rwxuca7x6y