Abstract
Reservoir computing, originally referred to as an echo state network or a liquid state machine, is a brain-inspired paradigm for processing temporal information. It involves learning a “read-out” interpretation for nonlinear transients developed by high-dimensional dynamics when the latter is excited by the information signal to be processed. This novel computational paradigm is derived from recurrent neural network and machine learning techniques. It has recently been implemented in photonic hardware for a dynamical system, which opens the path to ultrafast brain-inspired computing. We report on a novel implementation involving an electro-optic phase-delay dynamics designed with off-the-shelf optoelectronic telecom devices, thus providing the targeted wide bandwidth. Computational efficiency is demonstrated experimentally with speech-recognition tasks. State-of-the-art speed performances reach one million words per second, with very low word error rate. Additionally, to record speed processing, our investigations have revealed computing-efficiency improvements through yet-unexplored temporal-information-processing techniques, such as simultaneous multisample injection and pitched sampling at the read-out compared to information “write-in”.
3 More- Received 30 January 2015
DOI:https://doi.org/10.1103/PhysRevX.7.011015
Published by the American Physical Society under the terms of the Creative Commons Attribution 3.0 License. Further distribution of this work must maintain attribution to the author(s) and the published article’s title, journal citation, and DOI.
Published by the American Physical Society
Viewpoint
Reservoir Computing Speeds Up
Published 6 February 2017
A brain-inspired computer made with optoelectronic parts runs faster thanks to a hardware redesign, recognizing simple speech at the rate of 1 million words per second.
See more in Physics
Popular Summary
When dealing with complex problems such as image or speech recognition, traditional computers are limited in terms of computational efficiency and energy consumption. Computers today use an architecture where memory is physically separated from the processing unit, and instructions are executed step by step. Processing temporal information as the brain does, however, mixes memory and processing to achieve higher computational efficiency and flexibility. Implementing a brain-inspired computer in photonic hardware, which processes information via light, can lead to further improvements by making use of low-power optical components and providing high speed through the use of broadband telecom devices. We demonstrate the capabilities of such a device, based on a brain-inspired paradigm known as reservoir computing. Our approach exhibits state-of-the-art speed on speech recognition tasks, identifying up to one million words per second with very low error rates.
Our design uses off-the-shelf components to implement a reservoir computing architecture that relies on electro-optical phase delay dynamics, which encodes information in the phase of light waves, as opposed to their intensity, to provide more accurate and faster processing. We demonstrate speech recognition using a standard database of recorded words, which can be processed and identified by our system after a learning procedure. In addition to our speed performance, we find improvements in computing efficiency compared to other recent implementations of photonic reservoir computing.
Currently the physical parameters of this architecture need to be fine-tuned for the task at hand; a truly universal machine requires no such optimization. Improvements in the complexity of our architecture could achieve not only greater universality but also more computational power. Another big challenge is unsupervised learning, where there is no information about what the result should be.