Fig. 1: Introduction to neuromorphic computing.

a Illustration showing the sequence of three firing neurons (A, B, C) in the brain, and their synaptic connections. The axon terminal of one neuron communicates with the dendrites of the other neuron through a small synaptic gap. b Operating power consumption of a circuit as a function of clock frequency. Dynamic power dissipation is lower for low-frequency neuromorphic circuits that mimic the neurons and synapses in the brain. c Diagram (left) illustrating the mechanism of a neuron firing in an integrate and fire (IF) model where buildup of its action potential past the firing threshold (V_T) causes it to fire. IF circuitry (center) implementing neuron firing is implemented by a feedback loop adder and comparator circuit (with a firing threshold of N), while neural learning (right) through synaptic behavior implementation is accomplished by a subtractor and a counter circuit (C⁺ refers to an up-counter with Qs being the output). The neural learning circuit evaluates the synaptic weight between two firing neurons, allowing for the eventual training of the NM circuit.