High-Speed Photonic Reservoir Computing Using a Time-Delay-Based Architecture: Million Words per Second Classification

Abstract

Reservoir computing, originally referred to as an echo state network or a liquid state machine, is a braininspired 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 temporalinformation-processing techniques, such as simultaneous multisample injection and pitched sampling at the read-out compared to information "write-in".

Human action recognition with a large-scale brain-inspired photonic computer

“…Furthermore, as less parameters are inferred during training, the network can be trained on significantly smaller datasets without the risk of overfitting. The performance of the numerous experimental implementations of reservoir computing in electronics 8 , optoelectronics [9][10][11][12] , optics [13][14][15][16] , and integrated on chip 17 is comparable to other digital algorithms on a series of benchmark tasks, such as wireless channel equalisation 5 , phoneme recognition 18 and prediction of future evolution of financial time series 19 . Finally, it was shown that the readout layer of photonic reservoir computers can be implemented optically and trained using a digital micro-mirror device 20 .…”

Using machine learning to assess short term causal dependence and infer network links

“…where δ ij is the Kronecker delta, and we propose to use Eqs. (1) and (4) Although we use a specific reservoir computing implementation, we expect that, with suitable modifications, our approach can be adapted to 'deep' types of machine learning [2] , as well as to other implementations of reservoir computing [24,25,37,38] , (notably implementations involving photonics [24] , electronics [37] and field programmable gate arrays(FPGAs) [25] ). The input-to-reservoir coupling matrix W in couples the input time series for the vector z to the reservoir state vector r. The reservoir-to-output coupling matrix W out generates the output vectorẑ from the reservoir.ẑ is found to be a good estimate of z after training.…”

Large-Scale Spatiotemporal Photonic Reservoir Computer for Image Classification

“…Because of its structural simplicity, reservoir computers have motivated many hardware implementations ranging from electronics [14], [15] and spintronics [16], [17], to photonics [18], [19] with the goal of reaching ultra-fast processing speed with an energy consumption at least two order of magnitudes below that of a software-based RC running on a computer. Unprecedented classification speed have been recently achieved on the spoken-digit recognition task from the TIMIT46 data base with real-time processing speed ranging from 300,000 to 1,000,000 words analyzed per second using laser with optical feedback [19] and optoelectronic oscillators arXiv:2004.02542v1 [cs.NE] 6 Apr 2020 [20], respectively.…”