Brain-Inspired Photonic Computer

Main UBFC partners: FEMTO-ST (optics and Energy departements), Labex ACTION (chair of excellence Nadia Steiner), Laboratoire de Mathématiques de Besançon, FC Lab, INSERM

One of the world’s fastest photonic computer is made by FEMTO-ST.
Digital electronics and standard computers have reached their limits in dealing with complex tasks, such as dynamical data classification, processing information at ultra-high speed and with energy efficiency.
To address these challenges, FEMTO-ST has been designing and implementing a photonic processing system based on the Reservoir computing principle and which is able to provide a high speed/efficiency computational power by using standard optical telecom bandwidth.

Reservoir Computing (RC) refers to a new discipline -at the interface of physics, computer science and engineering – which studies and develops calculation units based on artificial neural networks, and which operation is schematically inspired by biological neurons. Some exciting results have recently appeared, including the experimental realization of reservoirs with state of the art performances for tasks, such as speech recognition or time series prediction.

Concept of photonic crystal reservoir computing Experimental setup made by FEMTO-ST

Concept of photonic crystal reservoir computing Experimental setup made by FEMTO-ST

Spoken digit recognition: thanks to its demonstrator, FEMTO-ST has succeeded in processing 1 million spoken digit per second with a Word Error Rate (WER) < 0.02% at 20GS/s. !

Targeted applications and socio-economic challenges:
Many practical complex problems that cannot be addressed with conventional digital computers could be dealt by the neuromorphic processor of FEMTO-ST, such as:

  • Medical imaging, large-scale physical and system modeling,
  • Industrial strategic decision making, advanced time series prediction tasks,
  • Forecasting of catastrophes: critical meteorologic events, earthquakes, tsunami, financial bubble crisis, etc.

Concretely, this smart technology is currently explored in a strongly interdisciplinary collaborative framework and for multiple applications:

  • Energy : to address accurate and preventive diagnostic of Fuel Cell stacks (partner: the Energy Dept at FEMTO-ST, see after),
  • Economy: to investigate the forecasting of real stock exchange time series data (with the Mathematics Laboratory of Besançon),
  • Health: to analyze high complexity electroencephalogram (EEG) signals (with a medical research group of the INSERM),
  • Nano-manufacturing: for the real-time control of a femto-second laser beam devoted to the next generation of the nano- and micro-machining technologies (with the Optics Dept. at FEMTO-ST)

Next step: a neuromorphic photonic chip developed by the Labex ACTION
Dedicated to the development of integrated smart systems, the Labex ACTION aims to achieve within the next years a miniaturized neuromorphic photonic processor. These works are conducted by Laurent Larger, manager of this ambitious demonstrator project, including the support of 1 PhD (Bogdan Penkovsky, see picture) and a chair of excellence (Nadia Steiner).

Zoom on the Prognostic and Health Management (PHM) of Fuel Cell Systems

UBFC partners : Optics and Energy departments of FEMTO-ST (including the chair of excellence Nadia Steiner “Fault Tolerant Control of Fuel Cell Systems), FCLAB (FR CNRS 3539) – www.fclab.fr

Fuel Cell Systems (FCS) have gained an outstanding position in the panorama of “green” energy conversion systems. Among their main technological challenges, long durability, reliability and optimal working operations of both stack and auxiliaries are issues of primary importance.

Solving these problems will give a great opportunity for a significant breakthrough towards mass diffusion (stationary, transportation, portable power) of the current FCS technology.

The teams involved in the Labex ACTION strive to provide an advanced diagnostic and even prognostic tool, which would allow an efficient management and improvement of the FCS lifetime. Indeed, the photonic neuromorphic processor is expected to perform real-time analysis of multiple external variables (e.g. current and voltage) so that it would be able to detect accurately and early enough many potential faults of the FCS due to internal chemical, thermal and electrical effects.

Contact: laurent.larger@femto-st.fr

Lastest UBFC publications linked to this topic and optical achievements:

Laser chimeras as a paradigm for multistable patterns in complex systems
Larger, Laurent; Penkovsky, Bogdan; Maistrenko, Yuri
NATURE COMMUNICATIONS Volume: 6 Article Number: 7752 Published: JUL 2015 (IF: 11.5)

Giant thermo-optical relaxation oscillations in millimeter-size whispering gallery mode disk resonators
Diallo, Souleymane; Lin, Guoping; Chembo, Yanne K.
OPTICS LETTERS  Volume: 40   Issue: 16   Pages: 3834-3837   Published: AUG 15 2015

Multimode Brillouin spectrum in a long tapered birefringent photonic crystal fiber
Tchahame, Joel Cabrel; Beugnot, Jean-Charles; Kudlinski, Alexandre; et al.
OPTICS LETTERS (Volume: 40   Issue: 18   Pages: 4281-4284   Published: SEP 15 2015)

Other references:
neuromorph2

  • Photonic information processing beyond Turing: an optoelectronic implementation of reservoir computing (Optics Express 20 (3), 3241 [2012]), L. Larger, M. C. Soriano, D. Brunner, L. Appeltant, J. M. Gutierrez, L. Pesquera, C. R. Mirasso, I. Fischer
  • Photonic Nonlinear Transient Computing with Multiple-Delay Wavelength Dynamics (Phys. Rev. Lett. 108, 244101 [2012]), R. Martinenghi, S. Rybalko, M. Jacquot, Y. Chembo, L. Larger,
  • Virtual Chimera States for Delayed-Feedback Systems (cover page of Phys. Rev. Lett. 111, 054103 [2013]), L. Larger, B. Penkovsky, and Y. Maistrenko

Key project: EU project PHOCUS (Towards a Photonic liquid state machine based on delay-coupled system) on fast photonic systems (http://ifisc.uib-csic.es/phocus/)

See also : Workshop “Dynamical systems and brain-inspired information processing” / 2 and 3 november 2015, Besançon (FEMTO-ST, Laboratory of Mathematics of Besançon)