Fault-Tolerant Spiking Neural Network Mapping Algorithm and Architecture to 3D-NoC-Based Neuromorphic Systems

Neuromorphic computing uses spiking neuron network models to solve machine learning problems in a more energy-efficient way when compared conventional artificial neural networks. However, mapping the various components neuromorphic hardware is not trivial realize desired model for an actual simulation. Moreover, neurons and synapses could be affected by noise due external interference or random actions of other (i.e., neurons), which eventually lead unreliable results. This work proposes fault-tolerant algorithm architecture 3D network-on-chip (NoC)-based system (R-NASH-II) based on rank selection mechanism (RSM). The RSM allows ranking rapid mapping. Evaluation results show that with our proposed mechanism, we maintain efficiency 100% 20% spare rate fault (40%) than previous framework. Monte Carlo simulation evaluation reliability shows has increased mean time failure (MTTF) technique 43% average. Furthermore, operational availability $4\times 4\times 4$ (smallest) notation="LaTeX">$6\times 6\times 6$ (largest) NoC 88% 67% respectively.