Our latest paper on magnetic devices for neuromorphic computing is now published in Applied Physics Letters! We show that the domain wall-magnetic tunnel junction device can act as a stochastic neuron, which is robust to recognizing noisy images compared to ideal leaky, integrate, and fire neurons. This shows how magnetic devices could be beneficial for noisy, edge computing. The neurons are made on the same chip as magnetic multi-weight synapses, showing both neuromorphic building blocks can be made together.
Thomas Leonard took the data with Harrison Jin and Sam Liu who led the neural network simulations; the research is supported by the National Science Foundation.
Click here to learn more.
