Between the 13th and the 16th of September, Bordeaux hosted ECOC’2021, the largest conference on optical communications in Europe. NEoteRIC had an important role in this conference through the participation of three partners of the project.
Prof. Jose Capmany, coordinator of the project and Head of the Photonics Research Labs at the UPV, participated in the workshop about post Moore Data Center Networks. This workshop focused on future cloud network technologies from subcomponent to architecture level to address the technological bottleneck of an incremental evolutionary development. Disruptive concepts for optics, electronics in the context of a long-term cloud evolution were analyzed and new cloud application drivers for the Post Moore data center were discussed. Prof. Capmany unveiled what are the benefits of optical computing vs. analog electrical processing and explained how optical switching will finally be able to disrupt data center networks given the rapid scale out of AI applications.
Prof. Peter Bienstman, from Ghent University-IMEC, co-organized a workshop on Neuromorphic Computing. This workshop aimed to investigate information processing based on neuromorphic circuits in the optical and electrical domain. It shed light on how and when such approaches may be ready for implementation into optical transmission systems and what advantages they may offer. Furthermore, current limits and technological hurdles were discussed. Prof. Charis Mesaritakis, from the University of the Aegean, addressed Neuromorphic schemes for next generation telecommunications and security applications. Conventional signal processing in the electrical domain based on binary computing faces various technological and economical limitations. To increase signal processing speeds and significantly lower energy consumption per bit, radically new paths have to be followed. In that respect concepts borrowed or adapted from nature and implemented in the optical domain, such as neural networks based on reservoir computing, are appealing. Other concepts such as photonic neuron architectures have been investigated in the past and offer radically new ways of information processing. Furthermore, bio inspired neural networks in the electrical domain may offer significant advantages such as much lower power consumption than today’s electrical signal processing circuits.
Finally, both Prof. Mesaritakis and Prof. Bienstman participated in a technical session dedicated to Optical Neural networks to talk about “Photonic Reservoir Computing based on Optical Filters in a Loop as a High Performance and Low-Power Consumption Equalizer for 100 Gbaud Direct Detection Systems” and “Experimental Demonstration of Nonlinear Fibre Distortion Compensation with Integrated Photonic Reservoir Computing” respectively.