d79c8788088c2193f0244d8f1f36d2db-Supplemental.pdf

We modified the original60,000/10,000 train/test split to50,000/10,000/10,000 train/validate/test split, by partitioning away the last10,000training samples to the validation set. First, NALSM exhibited coexistence of small and large synaptic weights, which is necessary for chaotic activity in spiking networks [5]. Asexpected, spiketrain autocorrelation functions of input neurons remained flat showing no decay with respect to lag time. For N-MNIST, the average was 3,033,045.40