Figure 5.

(a) Fully connected ANN topology showing 3 interconnected layers, namely, the input layer, the hidden layer and the output layer²⁶, (b) Schematic of potential integration of an integrate-and-fire neuron in a spiking neural network framework consisting of bipolar weights. The positive and negative weighted sums are computed using two separate dot-product engines and input to two different ring-resonators. The bidirectional integrating action of the two ports of the ring resonator is leveraged to calculate the effective membrane potential under the action of the bipolar weighted sums. Output spikes are generated when the effective membrane potential of the neuron crosses a threshold by the spike generation mechanism described. (c) The behavior of the proposed integrate-and-fire neuron in the simulated SNN showing the variation of the membrane potential under the action of incident pulses thus showing integrate and firing action.