High Speed Photonic Neuromorphic Computing Using Recurrent Optical Spectrum Slicing Neural Networks

Neuromorphic Computing implemented in photonic hardware is one of the most promising routes towards achieving machine learning processing at the picosecond scale, with minimum power consumption. In this work, we present a new concept for realizing photonic recurrent neural networks and reservoir computing architectures with the use of recurrent optical spectrum slicing. This is accomplished through simple optical filters placed in an loop, where each filter processes a specific spectral slice of the incoming optical signal. The synaptic weights in our scheme are equivalent to filters central frequencies and bandwidths. This new method for implementing recurrent neural processing in the photonic domain, which we call Recurrent Optical Spectrum Slicing Neural Networks, is numerically evaluated on a demanding, industry-relevant task such as high baud rate optical signal equalization 100 Gbaud, exhibiting ground-breaking performance. The performance enhancement surpasses state-of-the-art digital processing techniques by doubling the reach while minimizing complexity and power consumption by a factor of 10 compared to state-of-the-art solutions. In this respect, ROSS-NNs can pave the way for the implementation of ultra-efficient photonic hardware accelerators tailored for processing high-bandwidth optical signals in optical communication and high-speed imaging applications