Programmable surface plasmonic neural networks for microwave detection and processing

A range of alternative approaches to traditional digital hardware have been explored for the implementation artificial neural networks, including optical networks and diffractive deep networks. Spoof surface plasmon polariton waveguides, which operate at microwave terahertz frequencies, can offer low crosstalk, radiation loss easy integration, are potential use in development an technology Here, we report a programmable plasmonic network that is based on spoof platform detect process microwaves. The approach uses parallel coupled cell integrated with varactors. weight coefficients be adjusted by tuning voltages varactors, activation function programmed detecting input intensity feeding back threshold amplifier. We show two-layer fully connected consisting four cells output perform vector classification task. also used create wireless communication system decode recover images. In addition, partially classify handwritten digits high accuracy. weights functions.