arXiv subjects

Akram Akrout

Publications and source records attributed to Akram Akrout.

3 recordsLinked to original sources

arXiv2022

Photonic reservoir computer based on frequency multiplexing

Reservoir computing is a brain inspired approach for information processing, well suited to analogue implementations. We report a photonic implementation of a reservoir computer that exploits frequency domain multiplexing to encode neuron states. The system processes 25 comb lines simultaneously (i.e. 25 neurons), at a rate of 20 MHz. We illustrate performances on two standard benchmark tasks: channel equalization and time series forecasting. We also demonstrate that frequency multiplexing allows output weights to be implemented in the optical domain, through optical attenuation. We discuss the perspectives for high speed high performance low footprint implementations.

physics.optics↗

arXiv2016

Parallel photonic reservoir computing using frequency multiplexing of neurons

Today's unrelenting increase in demand for information processing creates the need for novel computing concepts. Reservoir computing is such a concept that lends itself particularly well to photonic hardware implementations. Over recent years, these hardware implementations have gained maturity and now achieve state-of-the-art performance on several benchmark tasks. However, implementations so far are essentially all based on sequential data processing, leaving the inherent parallelism of photonics unexploited. Parallel implementations process all neurons simultaneously, and therefore have the potential of reducing computation time by a factor equal to the number of neurons, compared to sequential architectures. Here, we report a parallel reservoir computer that uses frequency domain multiplexing of neuron states. We illustrate its performance on standard benchmark tasks such as nonlinear channel equalization, the reproduction of a nonlinear 10th-order system, and speech recognition, obtaining error rates similar to previous optical experiments. The present experiment is thus an important step towards high speed, low footprint, all optical photonic information processing.

cs.ET↗

arXiv2014

Analog input layer for optical reservoir computers

Reservoir computing is an information processing technique, derived from the theory of neural networks, which is easy to implement in hardware. Several reservoir computer hardware implementations have been realized recently with performance comparable to digital implementations, which demonstrated the potential of reservoir computing for ultrahigh bandwidth signal processing tasks. In all these implementations however the signal pre-processing necessary to efficiently address the reservoir was performed digitally. Here we show how this digital pre-processing can be replaced by an analog input layer. We study both numerically and experimentally to what extent the pre-processing can be replaced by a modulation of the input signal by either a single sine-wave or by a sum of two sine functions, since harmonic oscillations are particularly easy to generate in hardware. We find that the modulation by a single sine gives performance worse than state of the art. On the other hand, on many -but not all- tasks, the modulation by two sines gives performances comparable to the state of the art. The present work thus represents an important step towards fully autonomous, ultrahigh bandwidth analog reservoir computers.

cs.ET↗