Skip to main content
. 2018 Aug 3;12:435. doi: 10.3389/fnins.2018.00435

Figure 3.

Figure 3

Illustration of spike forward and backward propagation of a multi-layer SNN consisting of LIF neurons. In forward pass, the spiking neuron integrates the input current (net) generated by the weighted sum of the pre-neuronal spikes with the interconnecting synaptic weights and produces an output spike train. In backward pass, the derivatives of designated loss function with respect to each synaptic weight are calculated from chain-rule.