Regensburg 2016 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

DY: Fachverband Dynamik und Statistische Physik

DY 30: Poster - Complex nonlinear systems

DY 30.11: Poster

Dienstag, 8. März 2016, 18:15–21:00, Poster C

Reservoir computing simulated by using the Lang Kobayashi laser equations with time delayed feedback — •David Schicke, André Röhm, and Kathy Lüdge — Institut f. Theo. Physik, Sekr. EW 7-1, Technische Universität Berlin, Hardenbergstr. 36, 10623Berlin, Germany

Reservoir computing has the potential to outperform traditional van-Neumann architectures in terms of swiftness as well as accuracy. Traditionally achieved by a system consisting of about 200-1000 physical nodes which are connected to each other, reservoir computing can also be performed by a single node with time delayed self feedback. Physical nodes are simulated as virtual nodes along the delay line. To achieve different coupling strength between virtual nodes, the input also has to be masked. Two different non-linearities are compared, one being the Ikeda nonlinearity, already discussed in e.g. [1], [2] and the other the Lang Kobayashi differential equation describing laser dynamics with linewidth enhancement and optical feedback. Their computing capability will be tested with the NARMA10 task, their performance expressed by the normalized root mean square error.

[1] Larger, L. et al., Photonic information processing beyond turing: an optoelectronic implementation of reservoir computing, Optics Express 20, 3241-3249 (2012).

[2] Hermans, M. et al., Photonic Delay Systems as Machine Learning Implementations, arXiv: 1501.02592v1 [cs.NE] (2015)