The shape of light: how to measure, control and compute complexity? - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2019

The shape of light: how to measure, control and compute complexity?

Résumé

We review a selection of recent results where applications of concepts that harness the “shape of light” in spatial or temporal domain have been applied widely to yield significant advances in areas such as computer-vision sensing with digital holography, spatial shaping of complex laser beams and photonic neural networks.Computer vision is a powerful contact-less measurement tools successfully applied in numerous domains of application, where depth of field and working distances are constrained by the imaging magnification chosen. The use of pseudo-periodic patterns on the target of interest overcomes these usual computer-vision limitations leading to sub-pixel resolutions and making the absolute measurement range independent of the field-of-observation of the imaging system [1]. The approach was also validated using digital holography as imaging method with a tremendous enlargement of the allowed working distance range [2], and seems very well suited to diverse application needs in the micro-robotic and biomedical domains.Principles of digital holography can be used also for spatial shaping of complex laser beams such as Bessel, Airy or arbitrary beams using liquid crystal spatial phase modulators (SLM). Applications in micro&nano-machining by non-diffracting ultrashort laser pulses in various materials have been proposed during recent years [3]. This will also open new perspectives for applications of complex beams for applications in microscopy, optical coherence tomography or ultrafast physics.Photonic systems have revolutionized the hardware implementation of Recurrent Neural Networks and Reservoir Computing, in particular [4]. The fundamental principles of ReservoirComputing strongly facilitate a realization in such complex analog systems. Especially delay systems, which potentially provide large numbers of degrees of freedom even in simple architectures, can efficiently be exploited for information processing. We also demonstrated learning in large-scale neural networks with numerous nonlinear nodes in an architecture using SLM [5]. This last scheme is fully parallel and the passive weights maximize energy efficiency and bandwidth.In high-tech areas such as micro-robotics and photonics, measurement requirements are increasing in terms of high resolution and their controls are based on multi-scale and complex parameters. Increasingly real-time processing remains a big challenge for future applications, where next generation of systems will need to implement new hardware architectures, maybe based on photonic neural networks.
Fichier non déposé

Dates et versions

hal-02370475 , version 1 (13-05-2021)

Identifiants

  • HAL Id : hal-02370475 , version 1

Citer

Maxime Jacquot, Daniel Brunner, Javier Porte, Louis Andreoli, Sheler Maktoobi, et al.. The shape of light: how to measure, control and compute complexity?. 18th Workshop on Information Optics, Jul 2019, Stockholm, Sweden. ⟨hal-02370475⟩
60 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More