Authors: | T. Van Vaerenbergh, M. Fiers, J. Dambre, P. Bienstman | Title: | Towards integrated optical spiking neural networks: delaying spikes on chip | Format: | International Conference Proceedings | Publication date: | 12/2013 | Journal/Conference/Book: | Laser Dynamics and Nonlinear Photonics, 2013 Sixth "Rio De La Plata" Workshop on
(invited)
| Volume(Issue): | 1 p.9-12 | Location: | Montevideo, Uruguay | DOI: | 10.1109/LDNP.2013.6777418 | Citations: | 5 (Dimensions.ai - last update: 27/10/2024) 2 (OpenCitations - last update: 27/6/2024) Look up on Google Scholar
|
Abstract
To emulate Spiking Neural Networks (SNN) on an integrated photonic chip, one needs both
excitable optical components, to mimic the spiking neurons, and delay lines acting as the connections
that transfer the pulses between those neurons. Such an optical connection needs to have a delay in
the order of magnitude of the internal timescale of the neuron. Consequently, when using passive
Silicon-On-Insulator microrings as the basic excitable unit, those delay lines should have > 60 ns
delays, which is infeasible to obtain with standard integrated delay line approaches. In this paper,
we present in simulation a photonic neural circuit achieving a ∼ 200 ns spike delay. This paves the
way towards fully integrated optical SNNs. For instance, this delay line can be applied to build an
optical spike clock in the MHz range.
Related Research Topics
Related Projects
|
|