A Superconducting Nanowire Spiking Element for Neural Networks
November 12, (Thursday) 2020 11:30 am – 12:30 pm (EST)
In light of the growing need for faster, more energy-efficient computation, researchers are rapidly developing architectures inspired by the parallelism and performance of the human brain. Spiking neural networks are perhaps the most bio-realistic approach, mimicking the unique spiking dynamics of neurons to attain superior energy efficiency with the additional benefit of providing temporal information.
In this talk, Dr. Toomey presents a power-efficient artificial neuron made from superconducting nanowires, which naturally generates spiking based on the nonlinear transition between the superconducting and resistive states. Experimental results show that the device is capable of reproducing multiple bio-realistic behaviors, while simulations are used to explore how networks of nanowire neurons may be used as energy-efficient interference circuits or for studying theories of how biological neurons interact.
Dr. Emily Toomey currently works as a research scientist at the Laboratory for Physical Sciences (LPS) as part of their team that investigates novel qubit designs. Prior to joining LPS, Emily completed her PhD in Electrical Engineering at MIT as an NSF Graduate Research Fellow, where she developed memory cells and artificial neurons using superconducting nanowires. Her work seeks to understand and exploit the intrinsic nonlinearities in superconductors or other material systems in order to achieve new device functionality for energy-efficient computing. Before coming to MIT, Emily earned her B.S. in Electrical Engineering from Brown University, where she conducted an honors thesis that studied the enhanced conductivity of silver-intercalated poly-cytosine DNA using a scanning tunneling microscopy break-junction system. Outside of research, Emily is involved in science communication and journalism, and spends most of her time oil painting.