Skip to main content
. 2018 Mar 2;12:126. doi: 10.3389/fnins.2018.00126

Figure 9.

Figure 9

(A) Top-1/3/5 accuracy on 101-classes obtained with 8-training example (per class) learning for D/A models of different topologies. (B) Effect of variation in sparsity (for fixed 1 : 800 number of connections per neuron across different reservoir topologies) on the Top-1 accuracy achieved by the reservoir. Note, all results are performed on the D/A reservoir scheme with 8 training videos per class processed with the input spike transformation technique mentioned earlier.