Online Training of an Opto-Electronic Reservoir Computer

Antonik, Piotr; Duport, François; Smerieri, Anteo; Hermans, Michiel; Haelterman, Marc; Massar, Serge

doi:10.1007/978-3-319-26535-3_27

Cited by 17 publications

(17 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Typically, the training of the readout weights is performed off-line on a standard computer after the responses of the reservoir to the input examples have been recorded. An interesting approach to train on-line the readout weights is to use dedicated hardware such as field-programmable gate arrays (FPGAs) [63]. Although FPGAs are digital electronic devices, they are prepared to interact with analogue signals via on-board analogue-todigital and digital-to-analogue converters.…”

Section: Optoelectronic Delay-based Reservoir Computingmentioning

confidence: 99%

See 1 more Smart Citation

Advances in photonic reservoir computing

2017

View full text Add to dashboard Cite

Abstract:We review a novel paradigm that has emerged in analogue neuromorphic optical computing. The goal is to implement a reservoir computer in optics, where information is encoded in the intensity and phase of the optical field. Reservoir computing is a bio-inspired approach especially suited for processing time-dependent information. The reservoir's complex and high-dimensional transient response to the input signal is capable of universal computation. The reservoir does not need to be trained, which makes it very well suited for optics. As such, much of the promise of photonic reservoirs lies in their minimal hardware requirements, a tremendous advantage over other hardware-intensive neural network models. We review the two main approaches to optical reservoir computing: networks implemented with multiple discrete optical nodes and the continuous system of a single nonlinear device coupled to delayed feedback.

show abstract

Section: Optoelectronic Delay-based Reservoir Computingmentioning

confidence: 99%

“…The on-line training can then be realized by employing gradient descent or genetic algorithms. On-line learning capabilities offer the possibility to adapt to changing environments [63].…”

Section: Optoelectronic Delay-based Reservoir Computingmentioning

confidence: 99%

Advances in photonic reservoir computing

2017

View full text Add to dashboard Cite

show abstract

“…(a) Schematic representation of the simulated setup, based on the experimental system. 17 Optical and electronic components of the opto-electronic reservoir are shown in red and green, respectively. It contains an incoherent light source (SLED), a Mach-Zehnder intensity modulator (MZ), a 90/10 beam splitter, an optical attenuator (Att), an approximately 1.6 km fibre spool, two photodiodes (Pr and P f ), a resistive combiner (Comb) and an amplifier (Amp).…”

Section: Mackey-glass Chaotic Series Predictionmentioning

confidence: 99%

“…We have recently reported the first online-trained opto-electronic reservoir computer. 17 The key feature of this implementation is the FPGA chip, programmed to generate the input sequence, train the reservoir computer using the simple gradient descent algorithm, and compute the reservoir output signal in real time.…”

mentioning

confidence: 99%

Towards pattern generation and chaotic series prediction with photonic reservoir computers

et al. 2016

Self Cite

View full text Add to dashboard Cite

Reservoir Computing is a bio-inspired computing paradigm for processing time dependent signals that is particularly well suited for analog implementations. Our team has demonstrated several photonic reservoir computers with performance comparable to digital algorithms on a series of benchmark tasks such as channel equalisation and speech recognition. Recently, we showed that our opto-electronic reservoir computer could be trained online with a simple gradient descent algorithm programmed on an FPGA chip. This setup makes it in principle possible to feed the output signal back into the reservoir, and thus highly enrich the dynamics of the system. This will allow to tackle complex prediction tasks in hardware, such as pattern generation and chaotic and financial series prediction, which have so far only been studied in digital implementations. Here we report simulation results of our opto-electronic setup with an FPGA chip and output feedback applied to pattern generation and MackeyGlass chaotic series prediction. The simulations take into account the major aspects of our experimental setup. We find that pattern generation can be easily implemented on the current setup with very good results. The Mackey-Glass series prediction task is more complex and requires a large reservoir and more elaborate training algorithm. With these adjustments promising result are obtained, and we now know what improvements are needed to match previously reported numerical results. These simulation results will serve as basis of comparison for experiments we will carry out in the coming months.

show abstract

“…The optimal values of these weights are those for which the summation of all the different node responses always approaches the associated target as closely as possible. They can typically be determined with an offline training procedure using digital computers [5][6][7][8][11][12][13][14] or an online training procedure using a field-programmable gate array [17]. In our case, the training is done offline.…”

mentioning

confidence: 99%

Prediction performance of reservoir computing systems based on a diode-pumped erbium-doped microchip laser subject to optical feedback

et al. 2017

View full text Add to dashboard Cite

Reservoir computing (RC) systems are computational tools for information processing that can be fully implemented in optics. Here, we experimentally and numerically show that an optically pumped laser subject to optical delayed feedback can yield similar results to those obtained for electrically pumped lasers. Unlike with previous implementations, the input data are injected at a time interval that is much larger than the time-delay feedback. These data are directly coupled to the feedback light beam. Our results illustrate possible new avenues for RC implementations for prediction tasks. Reservoir computing (RC) is a brain-inspired concept for information processing that has been recently demonstrated to be efficient for solving practical time-dependent tasks [1,2]. RC systems operate by ensuring nonlinear mapping between the input and the output, thereby allowing a variety of information processing through training. To perform well, an RC system typically requires high dimensionality and nonlinearity. Traditionally, high dimensionality is obtained by randomly interconnecting a large number of neurons, while the needed nonlinearity can be implemented through sigmoidal activation functions. For example, with an ensemble of 16 interconnected semiconductor optical amplifiers, state-of-the-art performance has been achieved [3].Alternatively, state-of-the-art performance also has been obtained by relying on a single dynamical nonlinear node subject to delayed feedback [4]. This configuration (typically referred to as the delay-based RC) has the advantage of being easy to train and to implement experimentally. It has led to several implementations, even at high processing speeds, using stand-alone commercial telecommunication components [5][6][7][8]. The main differences between these experiments are the type of nonlinearity used and how the input matches with the period of the delay line. In those implementations, the nonlinear response of the reservoir is provided by passive nonlinearity such as saturable absorption of a semiconductor mirror [9][10][11] or by active devices such as optoelectronic modulators [5,8], optical amplifiers [3], or semiconductor lasers [7]. These experiments have been supported by numerical simulations [8,[12][13][14][15]. Numerical simulations also have shown that the different modes of multimode lasers subject to optical delayed feedback can be used independently to process independent tasks in parallel [16]. In all cases, the readout layer is trained (using some form of regression) from the state vectors of the reservoir in response to the training data.Until now, only one experiment has been dedicated to RC systems, in which the processing was done from the response provided by a laser subject to optical delayed feedback [7]. The laser used in this experiment was an electrically pumped singlelongitudinal-mode laser, and the input data were either electrically injected by modulating the pump current or optically injected into the reservoir through optical injection using another laser....

show abstract

Online Training of an Opto-Electronic Reservoir Computer

Cited by 17 publications

References 17 publications

Advances in photonic reservoir computing

Advances in photonic reservoir computing

Towards pattern generation and chaotic series prediction with photonic reservoir computers

Prediction performance of reservoir computing systems based on a diode-pumped erbium-doped microchip laser subject to optical feedback

Contact Info

Product

Resources

About