Published: July 26, 2017
Jacob Torrejon, Mathieu Riou, Flavio Abreu Araujo, Sumito Tsunegi, Guru Khalsa, Damien Querlioz, Paolo Bortolotti, Vincent Cros, K. Yakushiji, Akio Fukushima, Hitoshi Kubota, Shinji Yuasa, Mark D. Stiles, Julie Grollier
In the brain, hundred billions of neurons develop rhythmic activities and interact to process information. Taking inspiration from this behavior to realize high density, low power neuromorphic computing requires coupling huge numbers of nanoscale non-linear oscillators. However, there is no proof of concept today of neuromorphic computing with nanoscale oscillators. Indeed, nanodevices tend to be noisy and to lack the stability required to process data in a reliable way. Here, we show experimentally that, thanks to its well-controlled magnetization dynamics, a single nanoscale magnetic oscillator can recognize spoken digits at the state of the art compared to existing neural networks. We pinpoint the properties of the dynamical regime leading to highest performance in the bias conditions space. This result, combined with the exceptional ability of these magnetic oscillators to interact with other oscillators, their long lifetime, and low energy consumption, opens the path to fast parallel on-chip computations based on coupled networks of oscillators.
Pub Type: Journals
neuromorphic computing, reservoir computing, magnetic tunnel junction, spin-torque a nano-oscillator, artificial neurons
Created July 26, 2017, Updated November 10, 2018