Supervised learning of an artificial opto-magnetic neural network with picosecond laser pulses

The explosive growth of data and its related energy consumption is pushing the need to
develop novel, brain inspired and energy efficient schemes and materials for data processing
and storage. The prospect of continuous magnetization state manipulation of a Co/Pt thin film
via picosecond optical laser pulses offers the possibility of exploiting this material for neuro-
morphic computing. Here, we demonstrate that such material can be used as an artificial
adaptive synapse by controlling its internal magnetization state using circularly polarized
picosecond laser pulses. We also show experimentally an efficient implementation of
perceptron learning on a supervised opto-magnetic neural network, consisting out of such
magnetic synapses. Importantly, we show that optimization of synaptic weights is achieved by a
global feedback mechanism, such that learning does not rely on external storage or additional
optimization schemes.