1,720,990 research outputs found
Deep learning meets nanophotonics: A generalized Accurate predictor for near fields and far fields of arbitrary 3D nanostructures
No full textDeep artificial neural networks are powerful tools with many possible applications in nanophotonics. Here, we demonstrate how a deep neural network can be used as a fast, general purpose predictor of the full near-field and far-field response of plasmonic and dielectric nanostructures. A trained neural network is shown to infer the internal fields of arbitrary three-dimensional nanostructures many orders of magnitude faster compared to conventional numerical simulations. Secondary physical quantities are derived from the deep learning predictions and faithfully reproduce a wide variety of physical effects without requiring specific training. We discuss the strengths and limitations of the neural network approach using a number of model studies of single particles and their near-field interactions. Our approach paves the way for fast, yet universal, methods for design and analysis of nanophotonic systems
Deep learning enabled real time speckle recognition and hyperspectral imaging using a multimode fiber array
No full textWe demonstrate the use of deep learning for fast spectral deconstruction of speckle patterns. The artificial neural network can be effectively trained using numerically constructed multispectral datasets taken from a measured spectral transmission matrix. Optimized neural networks trained on these datasets achieve reliable reconstruction of both discrete and continuous spectra from a monochromatic camera image. Deep learning is compared to analytical inversion methods as well as to a compressive sensing algorithm and shows favourable characteristics both in the oversampling and in the sparse undersampling (compressive) regimes. The deep learning approach offers significant advantages in robustness to drift or noise and in reconstruction speed. In a proof-of-principle demonstrator we achieve real time recovery of hyperspectral information using a multi-core, multi-mode fiber array as a random scattering medium.</p
Dataset for Deep learning enabled real time speckle recognition and hyperspectral imaging using a multimode fiber array
No full textThis dataset supports the paper
Kürüm, U., Wiecha, P.R., French, R., & Muskens, O. L. (2018). Deep learning enabled real time speckle recognition and hyperspectral imaging using a multimode fiber array. Optics Express, DOI:10.1364/OE.27.020965</span
Dataset for Deep learning enabled strategies for modelling of complex aperiodic plasmonic metasurfaces of arbitrary size
No full textRaw data for numerical simulation results that support the paper
"Deep learning enabled strategies for modelling of complex aperiodic plasmonic metasurfaces of arbitrary size" published in ACS Photonics.
"dataset.zip" containing folders "figX" with X from 1 to 8
Each figure's dataset is bundled in an according subfolder. The data is labelled in an understandable form, if applicable headers have been added to raw-data textfiles.</span
Polarization conversion in plasmonic nanoantennas for metasurfaces using structural asymmetry and mode hybridization
No full textData in support of the paper by Peter R. Wiecha, Leo-Jay Black, Yudong Wang, Vincent Paillard, Christian Girard, Otto L. Muskens & Arnaud Arbouet entitled "Polarization conversion in plasmonic nanoantennas for metasurfaces using structural asymmetry and mode hybridization" in Scientific Reports 7, Article number: 40906 (2017) doi:10.1038/srep40906</span
Inverse design of unitary transmission matrices in silicon photonic coupled waveguide arrays using a neural adjoint model
No full textThe development of low-loss reconfigurable integrated optical devices enables further research into technologies including photonic signal processing, analogue quantum computing, and optical neural networks. Here, we introduce digital patterning of coupled waveguide arrays as a platform capable of implementing unitary matrix operations. Determining the required device geometry for a specific optical output is computationally challenging and requires a robust and versatile inverse design protocol. In this work we present an approach using high speed neural network surrogate-based gradient optimization, capable of predicting patterns of refractive index perturbations based on switching of the ultralow loss chalcogenide phase change material, antimony triselinide (Sb
2Se
3). Results for a 3 × 3 silicon waveguide array are presented, demonstrating control of both amplitude and phase for each transmission matrix element. Network performance is studied using neural network optimization tools such as data set augmentation and supplementation with random noise, resulting in an average fidelity of 0.94 for unitary matrix targets. Our results show that coupled waveguide arrays with perturbation patterns offer new routes for achieving programmable unitary operators, or Hamiltonians for quantum simulators, with a reduced footprint compared to conventional interferometer-mesh technology.
</p
Tailoring Second Harmonic Generation in single L-shaped plasmonic nanoantennas from the capacitive to conductive coupling regime
No full textReadme file with description of data set and full data set</span
Deep Learning for nanotechnology : crystal growth characterization and nano-photonics inverse design
No full textL'apprentissage profond, un sous-ensemble de l'intelligence artificielle, a considérablement amélioré les capacités des machines dans plusieurs domaines. L'apprentissage profond est basé sur l'architecture des réseaux neuronaux. Dans cette thèse, nous avons introduit les réseaux de neurones, leurs différentes architectures, leurs domaines d'application ainsi que le processus d'apprentissage, la descente de gradient stochastique en mini-batch. Nous avons ensuite utilisé des modèles de réseaux de neurones dans deux applications, à savoir la caractérisation de la croissanec des cristaux et la nanophotonique. Dans la première application, nous avons effectué trois tâches qui sont la détection de la désoxydation du substrat à l'aide d'un autoencodeur pour compresser les données, suivi d'un réseau de neurones convolutionnel pour classifier des séquences d'images compressées en oxydées et désoxydées, la classification des reconstructions de surface en (2 x 4) et c(4 x 4) et enfin la détermination de l'orientation du cristal par rapport au faisceau d'électrons du système RHEED à l'aide de réseaux de neurones résiduels. Dans la deuxième application, nous avons expliqué le problème de conception inverse en nanophotonique avant de mettre en œuvre certaines architectures d'apprentissage profond permettant de contourner ce problème inverse. La première architecture est le réseau tandem, un modèle génératif impliquant deux réseaux, la deuxième est l'autoencodeur variationnel qui est une variante de l'autoencodeur et la dernière architecture est la méthode dite de "neural adjoint", une méthode d'optimisation basée sur le gradient. À la fin de cette partie, nous proposons un réseau de neurones graphique qui permet de représenter les nanostructures comme des graphes. Enfin, nous avons testé ses capacités d'interpolation et d'extrapolation suivi d'un réglage fin pour inclure les données d'extrapolation dans le champ d'entraînement du modèle.Deep learning, a subset of artificial intelligence, has significantly improved machines capacities in several areas. Deep learning is based on its neural network architecture. In this thesis, we introduced neural networks, their different architectures, respective fields of use and the training process, the mini-batch stochastic gradient descent. We then used neural network models in two applications, namely RHEED image characterization and nanophotonics. In the first application, we performed three tasks, namely the detection of substrate deoxidation using an autoencoder to compress the data followed by a convolutional neural network to classify compressed image sequences into oxidized and deoxidized, the classification of surface reconstructions into (2 x 4) and c(4 x 4) and finally the determination of the crystal orientation with respect to the electron beam of the RHEED system using residual neural networks. In the second application, we explained the inverse design problem in nanophotonics before implementing some deep learning architectures designed to overcome the inverse problem. The first architecture is the tandem network, a generative model involving two networks, the second one is variational autoencoder which is a variant of the regular autoencoder and the last architecture is the neural adjoint method, a gradient-based optimization method. At the end of this part, we propose a graph neural network that allows nanostructures to be represented as graphs. At the end, We have tested its interpolation and extrapolation capabilities, with fintuning at the end to include the extrapolation data in the model's training field
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
- …
