LL Technology Office Seminar: Application-Tailored Neuromorphic Computing using Magnetic and 2D Materials

Neuromorphic computing promises efficient computing for artificial intelligence using co-design from materials through devices, circuits, systems, and applications. The device requirements vary drastically depending on application, e.g. radiation-tolerant circuits for edge computing in space, and bio-compatible materials for in-sensor computing for medicine and health. We will present on our results using magnetic spin textures to function as stochastic artificial neurons that are noise-resilient, multi-weight synapses that have high cycling stability, and leaky, integrate, and fire neurons that can mimic higher-order neuronal functions. We will show the developed devices are radiation tolerant for edge computing. We will also show how the magnetic devices can be tailored for probabilistic computing. Turning to neuromorphic computing for health, we will show our results building artificial neurons and synapses from bio-compatible graphene. These results show the promise of an across-the-stack approach to AI for hardware-aware computing.