par Antonik, Piotr 
;Hermans, Michiel ;Duport, Francois ;Haelterman, Marc ;Massar, Serge
Référence SPIE's 2016 Laser Technology and Industrial Laser Conference, Vol. 9732
Publication Publié, 2016-02-15
;Hermans, Michiel ;Duport, Francois ;Haelterman, Marc ;Massar, Serge Référence SPIE's 2016 Laser Technology and Industrial Laser Conference, Vol. 9732
Publication Publié, 2016-02-15
Publication dans des actes
| Résumé : | Reservoir Computing is a bio-inspired computing paradigm for processing time dependent signals that isparticularly well suited for analog implementations. Our team has demonstrated several photonic reservoircomputers with performance comparable to digital algorithms on a series of benchmark tasks such as channelequalisation and speech recognition. Recently, we showed that our opto-electronic reservoir computer could betrained online with a simple gradient descent algorithm programmed on an FPGA chip. This setup makes itin principle possible to feed the output signal back into the reservoir, and thus highly enrich the dynamics ofthe system. This will allow to tackle complex prediction tasks in hardware, such as pattern generation andchaotic and financial series prediction, which have so far only been studied in digital implementations. Herewe report simulation results of our opto-electronic setup with an FPGA chip and output feedback applied topattern generation and Mackey-Glass chaotic series prediction. The simulations take into account the majoraspects of our experimental setup. We find that pattern generation can be easily implemented on the currentsetup with very good results. The Mackey-Glass series prediction task is more complex and requires a largereservoir and more elaborate training algorithm. With these adjustments promising result are obtained, andwe now know what improvements are needed to match previously reported numerical results. These simulationresults will serve as basis of comparison for experiments we will carry out in the coming months. |
