Abstract
Reservoir computing is a recurrent machine learning framework that expands the dimensionality of a problem by mapping an input signal into a higher-dimension reservoir space that can capture and predict features of complex, non-linear temporal dynamics. Here, we report on a bulk electro-optical demonstration of a reservoir computer using speckles generated by propagating a laser beam modulated with a spatial light modulator through a multimode waveguide. We demonstrate that the hardware can successfully perform a multivariate audio classification task performed using the Japanese vowel speakers public data set. We perform full wave optical calculations of this architecture implemented in a chip-scale platform using an SiO2 waveguide and demonstrate that it performs as well as a fully numerical implementation of reservoir computing. As all the optical components used in the experiment can be fabricated using a commercial photonic integrated circuit foundry, our result demonstrates a framework for building a scalable, chip-scale, reservoir computer capable of performing optical signal processing.
Highlights
The rapid growth of data traffic leading to the emerging big data era has made special-purpose signal processing hardware a necessary tool
reservoir computing (RC) does not require backpropagation, which is computationally intensive. In this Article, we present results on spatially-distributed bulk optical RC using laser speckle in a multimode optical waveguide as the “reservoir”, which allows for parallelized input and could potentially be scaled into a photonic integrated circuit (PIC) to form a multi-GHz-bandwidth, analog photonic processor
In this Article, we demonstrated an optical reservoir computer using speckles generated from a multimode waveguide
Summary
The rapid growth of data traffic leading to the emerging big data era has made special-purpose signal processing hardware a necessary tool. Through optical waveguides and a 1D array of light modulators With this approach, we demonstrate that only 200 neurons are required to perform multivariate classification tasks. The calculations presented here demonstrate that a 100-micron wide, 10-cm long planar waveguide provides sufficient speckle mixing for building a 100-neuron reservoir computer Such planar waveguides can be fabricated in tight spirals using a silicon-oninsulator (SOI) PIC with a total footprint less than 1 cm2 [25, 26]. The processing speed of such a device would only be limited by the optical delay of the chip, the modulator bandwidth, and the photodiode response time This opens the possibility of performing real-time signal processing of radio-frequency (RF) and other high frequency signals, or rapid serial processing of very large datasets
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
