Skip to main content
. 2018 Aug 23;12:524. doi: 10.3389/fnins.2018.00524

Figure 6.

Figure 6

(A) Sparse representations of MNIST digits (28 × 28 in dimension) self-learned by a Liquid-SNN of 400 neurons (80% of which are excitatory) in the weight of the synapses connecting the input to the liquid-excitatory neurons (320 excitatory neurons organized in a 17 × 17 grid). (B) Sparse features self-learned by the individual liquids making up a two-liquid SpiLinC consisting of 200 neurons per liquid (160 excitatory neurons organized in a 12 × 12 grid) trained using the vertical partitions (each 28 × 16 in dimension) of the MNIST digit patterns.