Training a multilayer dynamical spintronic network with standard machine learning tools to perform time series classification

Ross et al. have experimentally demonstrated a multilayer network of spintronic oscillators, but with a feedforward architecture dedicated to static tasks [24]. The ability to process time-series (classification, prediction, Rodrigues et al. have shown by numerical simulations generation etc.) is important for many applications how to train the transient dynamics of a single layer network from smart sensors in industrial maintenance to of oscillators with optimal control theory, on a static personal assistants and medical devices. Using the dynamics task [25]. of a physical system, leveraging its non-linearity and memory for such processing has been widely explored Here we simulate and train a multi-layer network of with the development of recurrent neural networks, both spintronic oscillators as neurons, using standard machine from a purely mathematical perspective [1-4] as well as learning tools. We leverage the transient dynamics of the from a brain-inspired perspective with spiking recurrent oscillators to perform time-series classification of the sequential neural networks [5-8]. Chen et al. have shown that