Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz
AKPIK 3: AKPIK Postersession
AKPIK 3.1: Poster
Donnerstag, 30. September 2021, 13:30–15:30, P
Towards optical neuromorphic hardware at Cesium wavelength — •Elizabeth Robertson1,2, Mingwei Yang1, Luisa Esguerra1, Leon Meßner1, Lina Jaurigue2, Guillermo Gallego2,3, Kathy Lüdge2, and Janik Wolters1,2,3 — 1Deutsches Zentrum für Luft- und Raumfahrt, Institute for Optical Sensor Systems, Rutherfordstraße 2, 12489 Berlin, Germany — 2Technische Universität Berlin, Str. des 17. Junis 135, 10623 Berlin, Germany — 3Einstein Center Digital Future Robert-Koch-Forum, Wilhelmstraße 67, 10117 Berlin, Germany
With the exponential growth of research in Machine Learning, so too has the demand for fast, energy efficient neuromorphic hardware, to execute such algorithms. Optics provides an attractive tool for implementing neuromorphic hardware due to is speed, low crosstalk and high parallelism. In particular, convolutional- and recurrent neural networks benefit from an optics implementation; by using passive optics to carry out convolution [1], and time division multiplexing to demonstrate a reservoir computer [2]. We present our results towards realizing an optical convolutional neural network with an atomic non-linearity [3] and an all-optical reservoir computer based on a delay loop with an additional optical random-access memory (RAM)[4].
[1] Miscuglio, M. et al. Optica 7, 1812-1819 (2020).
[2] Brunner, D., et al. Nat Commun 4, 1364 (2013).
[3] Yang, M. et al. CLEO/Europe-EQEC, poster JSIV-P.4 (2021).
[4] Wolters, J., et al. Phys. Rev. Lett. 119, 060502 (2017).