Hybrid Spintronic-CMOS Spiking Neural Network with On-Chip Learning: Devices, Circuits, and Systems

It's hard to beat the brain at recognizing patterns, so neuromorphic computing platforms aim to mimic neuronal function with electronic hardware. The authors design a hybrid spintronic-electronic spiking neural network that exhibits online learning, at levels from device to system. Spintronic synapses offer the spike-timing-dependent plasticity needed for efficient, ultralow-power cognitive computing. This shows that the physics of spin-orbit torque can be used to engineer much more than just the next generation of data storage.