FPGA Implementation of Reservoir Computing with Online Learning
Piotr Antonik, Anteo Smerieri, François Duport, Marc Haelterman, Serge Massar · Dépôt institutionnel de l'Université libre de Bruxelles (Université Libre de Bruxelles) · 2015
Reservoir Computing is a bio-inspired computing paradigm for processing time dependent signals. The performance of its analogue implementation are comparable to, and sometimes surpass, other state of the art algorithms for tasks such as speech recognition or chaotic time series prediction. However, these implementation present several issues, which we address here by using programmable dedicated electronics in place of a personal computer. We demonstrate a standalone reservoir computer programmed onto a FPGA board and apply it to the real-world task of equalisation of a nonlinear communication channel. The training of the RC is carried out online, as this learning method, on top of being simple to implement, allows the RC to adapt itself to a changing environment, as we show here by equalising a variable communication channel.