Parallel photonic information processing at gigabyte per second data rates using transient states

A high-speed, optical, learning-based approach to information processing which uses components that are readily available in labs is presented in Nature Communications. The demonstration highlights the potential of photonic systems for high data-rate computation.

The ever-increasing demands on modern computer systems to perform complex calculations mean that more efficient processing methods are needed. One approach suggested to improve efficiency is that of reservoir computing, which uses a system of nonlinear transient states as its basis, much like a neural network. Daniel Brunner and colleagues implement a reservoir computing scheme using a photonic system built entirely using standard optical components. A semiconductor laser generates the transient states needed, with ordinary optical fibres enabling the computation. This type of system is suited to processing temporal information sequences, and the authors show that they can perform spoken digit and speaker recognition at gigabyte per second data rates.

This all-optical system demonstrates the possibilities of both reservoir computing and optical technology for efficient data processing tasks.

http://www.nature.com/ncomms/journal/v4/n1/full/ncomms2368.html

http://www.uib.es/noticies/LIFISC-CSIC-UIB-proposa-un-metode-de-processament.cid271659