Reservoir Computing (RCs) is a brain-inspired computational paradigm based on recurrent neural networks where only output weights are tuned, while internal weights remain untrained. Recently, a photonic RC system that encodes neurons in the lines of a frequency comb was demonstrated. A similar frequency-multiplexing approach was used to implement a single-layer feed-forward neural network. Here, we present the design for an integrated optical output layer for such frequency based photonic neural networks. The all-optical output layer uses wavelength (de)multiplexers and wavelength converters to apply signed weights to neurons encoded in comb lines. The design concept has been split in two integrated circuits comprehending a Silicon Photonic te
chnology for the weighting scheme and an Indium Phosphide platform for the non-linear optical operation. During this talk, simulations and design parameters will be presented and discussed in detail.
Zoom link: https://uibuniversitat.zoom.us/j/87008849910
Meeting ID: 870 0884 9910
