Towards reservoir computing with autonomous Boolean networks

Nicholas D. Haynes , David P. Rosin , Damien Rontani (1) , Daniel J. Gauthier (2)

1 LMOPS - Laboratoire Matériaux Optiques, Photonique et Systèmes
2 PROMES - Procédés, Matériaux et Energie Solaire

We present our preliminary work developing a reservoir computing platform using field-programmable gate arrays (FPGAs), with the ultimate goal of detecting features of complex systems. We show that the three basic properties required for reservoir computing, namely that different input states are mapped to different reservoir states, input states that are close together are mapped to identical reservoir states, and a fading memory, can be realized even in moderately-sized Boolean networks synthesized on an FPGA. The networks are realized as ring oscillators consisting of multiple-input XOR gates that accept an input (Boolean) voltage and time-delayed feedback. These oscillators exhibit long chaotic transients when the input voltage is flipped, and it is found that these transients offer promising dynamics for reservoir computing.

Domains: Engineering Sciences [physics]

Nicholas D. Haynes, David P. Rosin, Damien Rontani, Daniel J. Gauthier. Towards reservoir computing with autonomous Boolean networks. 33rd meeting of Dynamics Days US 2014, Jan 2014, Atlanta, United States. 2014. ⟨hal-01109808⟩