Publications

2021

• Reduced order model approach for imaging with waves
  
  • Borcea Liliana
  • Garnier Josselin
  • Mamonov Alexander
  • Zimmerling Jörn
  Inverse Problems, IOP Publishing, 2021, 38 (2), pp.025004. Abstract We introduce a novel, computationally inexpensive approach for imaging with an active array of sensors, which probe an unknown medium with a pulse and measure the resulting waves. The imaging function is based on the principle of time reversal in non-attenuating media and uses a data driven estimate of the ‘internal wave’ originating from the vicinity of the imaging point and propagating to the sensors through the unknown medium. We explain how this estimate can be obtained using a reduced order model (ROM) for the wave propagation. We analyze the imaging function, connect it to the time reversal process and describe how its resolution depends on the aperture of the array, the bandwidth of the probing pulse and the medium through which the waves propagate. We also show how the internal wave can be used for selective focusing of waves at points in the imaging region. This can be implemented experimentally and can be used for pixel scanning imaging. We assess the performance of the imaging methods with numerical simulations and compare them to the conventional reverse-time migration method and the ‘backprojection’ method introduced recently as an application of the same ROM. (10.1088/1361-6420/ac41d0)
  DOI : 10.1088/1361-6420/ac41d0
• Constraint-based Verification of Formation Control
  
  • Alexandre Dit Sandretto Julien
  • Chapoutot Alexandre
  • Garion Christophe
  • Thirioux Xavier
  • Ziat Ghiles
  , 2021, pp.7136-7141. (10.1109/CDC45484.2021.9683622)
  DOI : 10.1109/CDC45484.2021.9683622
• Direct-Sequence Spread Spectrum with Signal Space Diversity for High Resistance to Jamming
  
  • Arbi Tarak
  • Geller Benoit
  • Oudomsack Pierre
  , 2021. Jamming attacks can severely limit wireless networks availability and can cause serious damage, in particular for tactical applications. Over the past decades, Direct-Sequence Spread Spectrum (DSSS) has been used to enhance resistance to jamming. In this paper, we first analyze the performance of the DSSS modulation in the presence of malicious jamming; we take into account by considering different physical phenomena such as a large Doppler shift and we use at the receiver side robust synchronization algorithms. We then propose to consider jointly rotated constellations and the DSSS technique in order to enhance robustness against jamming, while keeping reasonable complexity. Simulations results underline the good performance of our proposal as it shows a gain of several dBs compared to the DSSS technique with conventional constellations.
• Static Analysis of ReLU Neural Networks with Tropical Polyhedra
  
  • Goubault Eric
  • Palumby Sébastien
  • Putot Sylvie
  • Rustenholz Louis
  • Sankaranarayanan Sriram
  , 2021, 12913, pp.166-190. This paper studies the problem of range analysis for feedforward neural networks, which is a basic primitive for applications such as robustness of neural networks, compliance to specifications and reachability analysis of neural-network feedback systems. Our approach focuses on ReLU (rectified linear unit) feedforward neural nets that present specific difficulties: approaches that exploit derivatives do not apply in general, the number of patterns of neuron activations can be quite large even for small networks, and convex approximations are generally too coarse. In this paper, we employ set-based methods and abstract interpretation that have been very successful in coping with similar difficulties in classical program verification. We present an approach that abstracts ReLU feedforward neural networks using tropical polyhedra. We show that tropical polyhedra can efficiently abstract ReLU activation function, while being able to control the loss of precision due to linear computations. We show how the connection between ReLU networks and tropical rational functions can provide approaches for range analysis of ReLU neural networks. We report on a preliminary evaluation of our approach using a prototype implementation. (10.1007/978-3-030-88806-0_8)
  DOI : 10.1007/978-3-030-88806-0_8
• EXplainable Neural-Symbolic Learning (X-NeSyL) methodology to fuse deep learning representations with expert knowledge graphs: The MonuMAI cultural heritage use case
  
  • Díaz-Rodríguez Natalia
  • Lamas Alberto
  • Sanchez Jules
  • Franchi Gianni
  • Donadello Ivan
  • Tabik Siham
  • Filliat David
  • Cruz Policarpo
  • Montes Rosana
  • Herrera Francisco
  Information Fusion, Elsevier, 2021. The latest Deep Learning (DL) models for detection and classification have achieved an unprecedented performance over classical machine learning algorithms. However, DL models are black-box methods hard to debug, interpret, and certify. DL alone cannot provide explanations that can be validated by a non technical audience such as end-users or domain experts. In contrast, symbolic AI systems that convert concepts into rules or symbols-such as knowledge graphs-are easier to explain. However, they present lower generalisation and scaling capabilities. A very important challenge is to fuse DL representations with expert knowledge. One way to address this challenge, as well as the performance-explainability trade-off is by leveraging the best of both streams without obviating domain expert knowledge. In this paper, we tackle such problem by considering the symbolic knowledge is expressed in form of a domain expert knowledge graph. We present the eXplainable Neural-symbolic learning (X-NeSyL) methodology, designed to learn both symbolic and deep representations, together with an explainability metric to assess the level of alignment of machine and human expert explanations. The ultimate objective is to fuse DL representations with expert domain knowledge during the learning process so it serves as a sound basis for explainability. In particular, X-NeSyL methodology involves the concrete use of two notions of explanation, both at inference and training time respectively: 1) EXPLANet: Expert-aligned eXplainable Part-based cLAssifier NETwork Architecture, a compositional convolutional neural network that makes use of symbolic representations, and 2) SHAP-Backprop, an explainable AI-informed training procedure that corrects and guides the DL process to align with such symbolic representations in form of knowledge graphs. We showcase X-NeSyL methodology using MonuMAI dataset for monument facade image classification, and demonstrate that with our approach, it is possible to improve explainability at the same time as performance. (10.1016/j.inffus.2021.09.022)
  DOI : 10.1016/j.inffus.2021.09.022
• Luminescent lanthanide nanoparticle-based imaging enables ultra-sensitive, quantitative and multiplexed <i>in vitro</i> lateral flow immunoassays
  
  • Mousseau F.
  • Féraudet-Tarisse C.
  • Simon S.
  • Gacoin T.
  • Alexandrou A.
  • Bouzigues C I
  Nanoscale, Royal Society of Chemistry, 2021, 13 (35), pp.14814 - 14824. Lateral Flow Assays (LFAs) have been extensively used on-site to rapidly detect analytes, possibly in complex media. However, standard gold nanoparticle-based LFAs lack sensitivity and cannot provide quantitative measurements with high accuracy. To overcome these limitations, we image lanthanidedoped nanoparticles (YVO 4 :Eu 40%) as new luminescent LFA probes, using a homemade reader coupled to a smartphone and propose an original image analysis allowing strip quantification regardless of the shape of the test band signal. This method is demonstrated for the detection of staphylococcal enterotoxins SEA, SEG, SEH, and SEI. A systematic comparison to state-of-the-art gold nanoparticle-based LFA revealed an analytical sensitivity enhancement of at least one order of magnitude. We furthermore provided measurements of absolute toxin concentration over two orders of magnitude and demonstrated simultaneous quantitative detection of multiple toxins with unaltered sensitivity. In particular, we reached concentrations 100 times lower than the ones reported in the literature for on-site multiplexed LFA targeting enterotoxins. Altogether, these results highlight that our luminescent nanoparticle-based method provides a powerful and versatile on-site framework to detect multiple biomolecules with sensitivity approaching that obtained by ELISA. This paves the way to a change of paradigm in the field of analytical immunoassays by providing fast in situ quantitative high sensitivity detection of biomarkers or pathogens. (10.1039/d1nr03358a)
  DOI : 10.1039/d1nr03358a
• New preconditioners for the Laplace and Helmholtz integral equations on open curves: analytical framework and numerical results
  
  • Alouges François
  • Averseng Martin
  Numerische Mathematik, Springer Verlag, 2021, 148 (2), pp.255-292. Helmholtz wave scattering by open screens in 2D can be formulated as first-kind integral equations which lead to ill-conditioned linear systems after discretization. We introduce two new preconditioners in the form of square-roots of on-curve differential operators both for the Dirichlet and Neumann boundary conditions on the screen. They generalize the so-called “analytical” preconditioners available for Lipschitz scatterers. We introduce a functional setting adapted to the singularity of the problem and enabling the analysis of those preconditioners. The efficiency of the method is demonstrated on several numerical examples. (10.1007/s00211-021-01189-5)
  DOI : 10.1007/s00211-021-01189-5
• Solidification of a rivulet: shape and temperature fields
  
  • Huerre Axel
  • Monier Antoine
  • Séon Thomas
  • Josserand Christophe
  Journal of Fluid Mechanics, Cambridge University Press (CUP), 2021, 914. The freezing of a water rivulet begins with a water thread flowing over a very cold surface, is naturally followed by the growth of an ice layer and ends up with a water rivulet flowing on a static thin ice wall. The structure of this final ice layer presents a surprising linear shape that thickens with the distance. This paper presents a theoretical model and experimental characterisation of the ice growth dynamics, the final ice shape and the temperature fields. In a first part, we establish a two-dimensional model, based on the advection–diffusion heat equations, that allows us to predict the shape of the ice structure and the temperature fields in both the water and the ice. Then, we study experimentally the formation of the ice layer and we show that both the transient dynamics and the final shape are well captured by the model. In a last part, we characterise experimentally the temperature fields in the ice and in the water, using an infrared camera. The model shows an excellent agreement with the experimental fields. In particular, it predicts well the linear decrease of the water surface temperature observed along the plane, confirming that the final ice shape is a consequence of the interaction between the thermal boundary layer and the free surface. (10.1017/jfm.2021.41)
  DOI : 10.1017/jfm.2021.41
• Multi-Paradigm Modeling for Cyber-Physical Systems: A Systematic Mapping Review
  
  • Barisic Ankica
  • Ruchkin Ivan
  • Savić Dušan
  • Abshir Mohamed Mustafa
  • Al-Ali Rima
  • Li Letitia W
  • Mkaouar Hana
  • Eslampanah Raheleh
  • Challenger Moharram
  • Blouin Dominique
  • Nikiforova Oksana
  • Cicchetti Antonio
  Journal of Systems and Software, Elsevier, 2021. Cyber-Physical Systems (CPS) are heterogeneous and require cross-domain expertise to model. The complexity of these systems leads to questions about prevalent modeling approaches, their ability to integrate heterogeneous models, and their relevance to the application domains and stakeholders. The methodology for Multi-Paradigm Modeling (MPM) of CPS is not yet fully established and standardized, and researchers apply existing methods for modeling of complex systems and introducing their own. No systematic review has been previously performed to create an overview of the field on the methods used for MPM of CPS. In this paper, we present a systematic mapping study that determines the models, formalisms, and development processes used over the last decade. Additionally, to determine the knowledge necessary for developing CPS, our review studied the background of actors involved in modeling and authors of surveyed studies. The results of the survey show a tendency to reuse multiple existing formalisms and their associated paradigms, in addition to a tendency towards applying transformations between models. These findings suggest that MPM is becoming a more popular approach to model CPS, and highlight the importance of future integration of models, standardization of development process and education. (10.1016/j.jss.2021.111081)
  DOI : 10.1016/j.jss.2021.111081
• Crushing of additively manufactured thin-walled metallic lattices: Two-scale strain localization analysis
  
  • Balit Yanis
  • Margerit Pierre
  • Charkaluk Eric
  • Constantinescu Andrei
  Mechanics of Materials, Elsevier, 2021, 160, pp.103915. The response of architectured structures is characterized by multi-scale kinematics, which complex relation and effect on the engineering load response is still not completely understood and therefore needs further investigations. More precisely, the lack of experimental methods enabling to provide multi-scale data remains a key issue. The paper presents an experimental and numerical analysis of crushing tests performed on thin-walled auxetic metallic lattices manufactured by Directed Energy Deposition. The work is focused on the two-scale strain localization occurring (a) at the microscopic scale of the unit cell and (b) at the macroscopic scale corresponding to the homogenized continuum. The structures of interest are defined as the extrusion of a 2D auxetic wireframe and allow the application of an adapted digital image correlation scheme dedicated to the identification of the kinematics at the two considered scales. In particular, the microscopic kinematics are studied by following the deformation of lattice crossings, while the macroscopic strains are deduced from the motion of virtual unit cell corners. The results show that the global elastic-plastic response of the lattices is completely driven by the formation of plastic hinges at specific locations, leading to characteristic deformation patterns and eventually a collective behavior of neighboring unit cells. Companion finite element computations show an excellent match with experiments and thus enable to assess the effect of modeling assumptions, unit cell geometry, strain rate and geometrical imperfections in the global response of the architecture material. (10.1016/j.mechmat.2021.103915)
  DOI : 10.1016/j.mechmat.2021.103915
• Imaging in Random Media by Two-Point Coherent Interferometry
  
  • Garnier Josselin
  • Borcea Liliana
  SIAM Journal on Imaging Sciences, Society for Industrial and Applied Mathematics, 2021, 14 (4), pp.1635-1668. (10.1137/21M142068X)
  DOI : 10.1137/21M142068X
• Wave Propagation in Periodic and Random Time-Dependent Media
  
  • Garnier Josselin
  Multiscale Modeling and Simulation: A SIAM Interdisciplinary Journal, Society for Industrial and Applied Mathematics, 2021, 19 (3), pp.1190-1211. (10.1137/20M1377734)
  DOI : 10.1137/20M1377734
• Passive Communication with Ambient Noise
  
  • Garnier Josselin
  SIAM Journal on Applied Mathematics, Society for Industrial and Applied Mathematics, 2021, 81 (3), pp.814-833. (10.1137/20M1366848)
  DOI : 10.1137/20M1366848
• PARAFOG v2.0: A near-real-time decision tool to support nowcasting fog formation events at local scales
  
  • Ribaud Jean François
  • Haeffelin Martial P.A.
  • Dupont Jean Charles
  • Drouin Marc Antoine
  • Toledo Felipe
  • Kotthaus Simone
  Atmospheric Measurement Techniques, European Geosciences Union, 2021, 14 (12), pp.7893-7907. An improved version of the near-real-time decision tool PARAFOG (PFG2) is presented to retrieve pre-fog alert levels and to discriminate between radiation (RAD) and stratus lowering (STL) fog situations. PFG2 has two distinct modules to monitor the physical processes involved in RAD and STL fog formation and is evaluated at European sites. The modules are based on innovative fuzzy logic algorithms to retrieve fog alert levels (low, moderate, high) specific to RAD/STL conditions, minutes to hours prior to fog onset. The PFG2-RAD module assesses also the thickness of the fog. Both the PFG2-RAD and PFG2-STL modules rely on the combination of visibility observations and automatic lidar and ceilometer (ALC) measurements. The overall performance of the PFG2-RAD and PFG2-STL modules is evaluated based on 9 years of measurements at the SIRTA (Instrumented Site for Atmospheric Remote Sensing Research) observatory near Paris and up to two fog seasons at the Paris-Roissy, Vienna, Munich, and Zurich airports. At all sites, pre-fog alert levels retrieved by PFG2 are found to be consistent with the local weather analysis. The advanced PFG2 algorithm performs with a hit rate of about 100% for both considered fog types and presents a false alarm ratio on the order of 10% (30%) for RAD (STL) fog situations. Finally, the first high alerts that result in a subsequent fog event are found to occur for periods of time ranging from -120min to fog onset, with the first high alerts occurring earlier for RAD than STL cases. (10.5194/amt-14-7893-2021)
  DOI : 10.5194/amt-14-7893-2021