Design of Spiking Neural Networks (SNN) with Domain-Wall Devices

In this chapter, we extend the design of domain-wall-based neuromorphic devices from artificial neural networks (ANN), non-spiking in nature, to spiking neural networks (SNN), introduced in a previous chapter. We first show how a capacitor-charging-discharging-based analog circuit can be coupled to the same domain-wall device discussed in earlier chapters to make it implement spike-time-dependent plasticity (STDP), as opposed to just long-term potentiation (LTP) and long-term depression (LTD) like before. Next, we show how another analog circuit can be coupled to the same domain-wall device so that the device can act as a leaky integrate fire (LIF) neuron. Finally, we discuss the classification performance of an SNN based on these domain-wall-device-based neurons and synapses on different machine learning data sets.