Fig. 7: Experiments with spiking neural networks.
From: Introducing principles of synaptic integration in the optimization of deep neural networks

a Test accuracy for SNU and sSNU networks with three and two hidden layers, respectively. The results are mean and standard deviation of the test accuracy after training over five independent runs. b Test curves for SNU 3 × 256 networks with decreasing values for the learning rate η. Left hand-side: η = 0.01. Right hand-side: η = 0.001. The initial 100 training epochs are shown. c Test curves for sSNU networks with increasing number of hidden layers. Left hand-side: 4 × 256 networks. Right hand-side: 6 × 256 networks. In (b) and (c) the curves correspond to a single run.