Fig. 3: Comprehensive evaluation of MSF on the static and dynamic recognition tasks.
From: A multisynaptic spiking neuron for simultaneously encoding spatiotemporal dynamics
Performance-power comparison of different models and neurons on CIFAR-10 (a), and ImageNet (b) datasets. c Performance comparison of LIF, ReLU, and MSF on CIFAR10-DVS. d Event stream data of one sample from HARDVS and the corresponding accumulated frame images. e Performance-power comparison of LIF, ReLU, and MSF on HARDVS. f Spatial distribution of spike maps at four time points for one hidden layer of MSF and LIF neurons on HARDVS. g Synaptic number distributions of multisynaptic neurons in the human cerebral cortex28 and our trained models on the CIFAR-10, HARDVS, and ImageNet datasets.
