Publications

2023

• Optimal Trajectories of a UAV Base Station Using Hamilton-Jacobi Equations
  
  • Coupechoux Marceau
  • Darbon Jerome
  • Kelif Jean-Marc
  • Sigelle Marc
  IEEE Transactions on Mobile Computing, Institute of Electrical and Electronics Engineers, 2023, 22 (8), pp.4837 - 4849. We consider the problem of optimizing the trajectory of an Unmanned Aerial Vehicle (UAV). Assuming a traffic intensity map of users to be served, the UAV must travel from a given initial location to a final position within a given duration and serves the traffic on its way. The problem consists in finding the optimal trajectory that minimizes a certain cost depending on the velocity and on the amount of served traffic. We formulate the problem using the framework of Lagrangian mechanics. We derive closed-form formulas for the optimal trajectory when the traffic intensity is quadratic (single-phase) using Hamilton-Jacobi equations. When the traffic intensity is bi-phase, i.e. made of two quadratics, we provide necessary conditions of optimality that allow us to propose a gradient-based algorithm and a new algorithm based on the linear control properties of the quadratic model. These two solutions are of very low complexity because they rely on fast convergence numerical schemes and closed form formulas. These two approaches return a trajectory satisfying the necessary conditions of optimality. At last, we propose a data processing procedure based on a modified K-means algorithm to derive a bi-phase model and an optimal trajectory simulation from real traffic data. (10.1109/TMC.2022.3156822)
  DOI : 10.1109/TMC.2022.3156822
• Two-aperture measurements for GEO-feeder adaptive optics pre-compensation optimization
  
  • Lognoné Perrine
  • Conan Jean-Marc
  • Rekaya Ghaya
  • Paillier Laurie
  • Védrenne Nicolas
  Optics Letters, Optical Society of America - OSA Publishing, 2023, 48 (17), pp.4550. We present a method to estimate the pre-compensation phase of ground-to-geostationary orbit (GEO) optical links based on downlink phase and log-amplitude measurements from two ground apertures. This method allows us to reduce the point-ahead anisoplanatism that currently limits the telecom performance of GEO-feeder links. It is shown to reduce the anisoplanatic phase variance by 50%, hence improving the statistics of the coupled flux aboard the satellite. It also outperforms the one-aperture estimation method for very severe atmospheric conditions. Besides, only low-resolution amplitude measurements are required on the second aperture to reach the performance of the novel estimator. (10.1364/OL.495200)
  DOI : 10.1364/OL.495200
• The association between real-life markers of phone use and cognitive performance, health-related quality of life and sleep
  
  • Eeftens Marloes
  • Pujol Sophie
  • Klaiber Aaron
  • Chopard Gilles
  • Riss Andrin
  • Smayra Florian
  • Flückiger Benjamin
  • Gehin Thomas
  • Diallo Kadiatou
  • Wiart Joe
  • Mazloum Taghrid
  • Mauny Frédéric
  • Röösli Martin
  Environmental Research, Elsevier, 2023, 231, pp.116011. (10.1016/j.envres.2023.116011)
  DOI : 10.1016/j.envres.2023.116011
• Fully synthetic training for image restoration tasks
  
  • Achddou Raphaël
  • Gousseau Yann
  • Ladjal Saïd
  Computer Vision and Image Understanding, Elsevier, 2023, 233. In this work, we show that neural networks aimed at solving various image restoration tasks can be successfully trained on fully synthetic data. In order to do so, we rely on a generative model of images, the scaling dead leaves model, which is obtained by superimposing disks whose size distribution is scale-invariant. Pairs of clean and corrupted synthetic images can then be obtained by a careful simulation of the degradation process. We show on various restoration tasks that such a synthetic training yields results that are only slightly inferior to those obtained when the training is performed on large natural image databases. This implies that, for restoration tasks, the geometric contents of natural images can be nailed down to only a simple generative model and a few parameters. This prior can then be used to train neural networks for specific modality, without having to rely on demanding campaigns of natural images acquisition. We demonstrate the feasibility of this approach on difficult restoration tasks, including the denoising of smartphone RAW images and the full development of low-light images.
• Lessons for Interactive Theorem Proving Researchers from a Survey of Coq Users
  
  • de Almeida Borges Ana
  • Artís Annalí Casanueva
  • Falleri Jean-Rémy
  • Gallego Arias Emilio Jesús
  • Martin-Dorel Érik
  • Palmskog Karl
  • Serebrenik Alexander
  • Zimmermann Théo
  , 2023, 268 (12), pp.1-18. The Coq Community Survey 2022 was an online public survey of users of the Coq proof assistant conducted during February 2022. Broadly, the survey asked about use of Coq features, user interfaces, libraries, plugins, and tools, views on renaming Coq and Coq improvements, and also demographic data such as education and experience with Coq and other proof assistants and programming languages. The survey received 466 submitted responses, making it the largest survey of users of an interactive theorem prover (ITP) so far. We present the design of the survey, a summary of key results, and analysis of answers relevant to ITP technology development and usage. In particular, we analyze user characteristics associated with adoption of tools and libraries and make comparisons to adjacent software communities. Notably, we find that experience has significant impact on Coq user behavior, including on usage of tools, libraries, and integrated development environments. (10.4230/LIPIcs.ITP.2023.12)
  DOI : 10.4230/LIPIcs.ITP.2023.12
• Hybrid Energy Based Model in the Feature Space for Out-of-Distribution Detection
  
  • Lafon Marc
  • Ramzi Elias
  • Rambour Clément
  • Thome Nicolas
  , 2023. Out-of-distribution (OOD) detection is a critical requirement for the deployment of deep neural networks. This paper introduces the HEAT model, a new post-hoc OOD detection method estimating the density of in-distribution (ID) samples using hybrid energy-based models (EBM) in the feature space of a pre-trained backbone. HEAT complements prior density estimators of the ID density, e.g. parametric models like the Gaussian Mixture Model (GMM), to provide an accurate yet robust density estimation. A second contribution is to leverage the EBM framework to provide a unified density estimation and to compose several energy terms. Extensive experiments demonstrate the significance of the two contributions. HEAT sets new state-of-the-art OOD detection results on the CIFAR-10 / CIFAR-100 benchmark as well as on the large-scale Imagenet benchmark. The code is available at: https://github.com/MarcLafon/heatood.
• Integrating Prior Knowledge in Contrastive Learning with Kernel
  
  • Dufumier Benoit
  • Barbano Carlo Alberto
  • Louiset Robin
  • Duchesnay Edouard
  • Gori Pietro
  , 2023. Data augmentation is a crucial component in unsupervised contrastive learning (CL). It determines how positive samples are defined and, ultimately, the quality of the learnt representation. In this work, we open the door to new perspectives for CL by integrating prior knowledge, given either by generative models-viewed as prior representations-or weak attributes in the positive and negative sampling. To this end, we use kernel theory to propose a novel loss, called decoupled uniformity, that i) allows the integration of prior knowledge and ii) removes the negative-positive coupling in the original InfoNCE loss. We draw a connection between contrastive learning and conditional mean embedding theory to derive tight bounds on the downstream classification loss. In an unsupervised setting, we empirically demonstrate that CL benefits from generative models to improve its representation both on natural and medical images. In a weakly supervised scenario, our framework outperforms other unconditional and conditional CL approaches. Source code is available at this https URL.
• Robust Consensus in Ranking Data Analysis: Definitions, Properties and Computational Issues
  
  • Irurozki Ekhine
  • Goibert Morgane
  • Calauzènes Clément
  • Clémençon Stéphan
  , 2023. As the issue of robustness in AI systems becomes vital, statistical learning techniques that are reliable even in presence of partly contaminated data have to be developed. Preference data, in the form of (complete) rankings in the simplest situations, are no exception and the demand for appropriate concepts and tools is all the more pressing given that technologies fed by or producing this type of data (e.g. search engines, recommending systems) are now massively deployed. However, the lack of vector space structure for the set of rankings (i.e. the symmetric group S n) and the complex nature of statistics considered in ranking data analysis make the formulation of robustness objectives in this domain challenging. In this paper, we introduce notions of robustness, together with dedicated statistical methods, for Consensus Ranking the flagship problem in ranking data analysis, aiming at summarizing a probability distribution on S n by a median ranking. Precisely, we propose specific extensions of the popular concept of breakdown point, tailored to consensus ranking, and address the related computational issues. Beyond the theoretical contributions, the relevance of the approach proposed is supported by an experimental study.
• Exploring the Impact of Negative Sampling on Patent Citation Recommendation
  
  • Dessi Rima
  • Aras Hidir
  • Alam Mehwish
  , 2023. Due to the increasing number of patents being published every day, patent citation recommendations have become one of the challenging tasks. Since patent citations may lead to legal and economic consequences, patent recommendations are even more challenging as compared to scientific article citations. One of the crucial components of the patent citation algorithm is negative sampling which is also a part of many other tasks such as text classification, knowledge graph completion, etc. This paper, particularly focuses on proposing a transformer-based ranking model for patent recommendations. It further experimentally compares the performance of patent recommendations based on various state-of-the-art negative sampling approaches to measure and compare the effectiveness of these approaches to aid future developments. These experiments are performed on a newly collected dataset of US patents from Google patents. (10.5281/zenodo.7870197)
  DOI : 10.5281/zenodo.7870197
• Information theory as a unifying tool for understanding and designing human-computer interaction
  
  • Rioul Olivier
  , 2023. Information theory, particularly in a Bayesian context, has recently regained interest as a unified tool to understand and design human-computer communication and interaction. Like in everyday life, by continuously making predictions using previous experiences, the human-computer interaction should somehow aim at reducing the level of uncertainty (entropy), reinforcing true predictions and correcting wrong ones, via some feedback interaction loop. Taking the stance that human-computer interaction can be considered as a communication process, where uncertainty and information are described using information-theoretic terms, we can design interaction in such a way to optimize decisions under uncertainty. In my presentation I will illustrate these concepts in the context of a Bayesian experimental design using an expected utility function, where the computer can "runs experiments" on the user by sending feedback that maximizes the expected gain of information by the computer, and exploits the users’ subsequent input to update its knowledge as interaction progresses. Applications to Fitts' law, fast file retrieval and multiscale navigation are given to illustrate the concepts.
• A Compact, Quad-Band, and Wideband Antenna Using Triple-Band AMC
  
  • Gonçalves Licursi de Mello Rafael
  • Lepage Anne Claire
  • Begaud Xavier
  , 2023. We present a compact antenna with stable, unidirectional, high-gain radiation patterns to cover the standards 5G/4G/Wi-Fi 2.4/5/6E and X-band communications. The operating frequency bands are 2.4–2.7, 3.4–3.8, 5.17–6.45, and 8.0–12.0 GHz. The thickness is 0.09λl and the aperture size is 0.51×0.65 λl², where λl is the wavelength at the lowest frequency. To achieve the desired performance, we exploit both the in-band and out-of-band operation modes of a triple-band artificial magnetic conductor (AMC) illuminated by a grooved bow-tie radiating element surrounded by a metal ring.
• Design and measurement of an oblique wide-angle metamaterial absorber for RF space applications
  
  • Lepage Anne Claire
  • Begaud Xavier
  • Rance Olivier
  • Elis Kevin
  • Capet Nicolas
  , 2023. The paper deals with the design, realization and measurement of a lightweight electromagnetic metamaterial absorber for space applications operating on the [2 GHz, 2.3 GHz] frequency band, under oblique incidence from 35° to 65°.
• Fibrosis severity scoring on Sirius red histology with multiple-instance deep learning
  
  • Naik Sneha
  • Forlano Roberta
  • Manousou Pinelopi
  • Goldin Robert
  • Angelini Elsa
  Biological Imaging, Cambridge University Press, 2023, 3, pp.e17. Abstract Non-alcoholic fatty liver disease (NAFLD) is now the leading cause of chronic liver disease, affecting approximately 30% of people worldwide. Histopathology reading of fibrosis patterns is crucial to diagnosing NAFLD. In particular, separating mild from severe stages corresponds to a critical transition as it correlates with clinical outcomes. Deep Learning for digitized histopathology whole-slide images (WSIs) can reduce high inter- and intra-rater variability. We demonstrate a novel solution to score fibrosis severity on a retrospective cohort of 152 Sirius-Red WSIs, with fibrosis stage annotated at slide level by an expert pathologist. We exploit multiple instance learning and multiple-inferences to address the sparsity of pathological signs. We achieved an accuracy of $ 78.98\pm 5.86\% $ , an F1 score of $ 77.99\pm 5.64\%, $ and an AUC of $ 0.87\pm 0.06 $ . These results set new state-of-the-art benchmarks for this application. (10.1017/S2633903X23000144)
  DOI : 10.1017/S2633903X23000144
• Generation of non-classical light using semiconductor quantum dot lasers
  
  • Zhao Shiyuan
  • Ding Shihao
  • Huang Heming
  • Zaquine Isabelle
  • Belabas Nadia
  • Grillot Frédéric
  , 2023, pp.1-2. With a suppressed-pump-noise configuration, we have observed broadband amplitude-squeezed states of light generated from a 1310-nm distributed feedback quantum dot laser at room temperature. (10.1109/SUM57928.2023.10224418)
  DOI : 10.1109/SUM57928.2023.10224418
• APPLYING DEEP LEARNING TO P-BAND SAR TOMOGRAPHIC IMAGING IN PREPARATION FOR THE FUTURE BIOMASS MISSION
  
  • Berenger Zoé
  • Denis Loïc
  • Tupin Florence
  • Ferro-Famil Laurent
  , 2023. With Synthetic Aperture Radar tomography, it is possible to reconstruct reflectivity profiles in the direction orthogonal to the line-of-sight. When only a small number of interferometric baselines is available, the spatial resolution of profiles produced by beamforming is insufficient. While many iterative algorithms have been proposed in the past years to achieve improved tomographic reconstructions, these methods often require a large computational cost. In this paper we explore the use of a lightweight neural network to dramatically accelerate tomographic reconstruction in anticipation of the deluge of data generated by the future BIOMASS satellite. (10.1109/IGARSS52108.2023.10282400)
  DOI : 10.1109/IGARSS52108.2023.10282400
• DESPECKLING OF DUAL-POL GRD SENTINEL-1 IMAGES IN EXTRA-WIDE MODE BY DEEP LEARNING
  
  • Meraoumia Inès
  • Debanshu Ratha
  • Dalsasso Emanuele
  • Denis Loïc
  • Tupin Florence
  • Marinoni Andrea
  , 2023.
• Leveraging Vision-Language Foundation Models for Fine-Grained Downstream Tasks
  
  • Coquenet Denis
  • Rambour Clément
  • Dalsasso Emanuele
  • Thome Nicolas
  , 2023. Vision-language foundation models such as CLIP have shown impressive zero-shot performance on many tasks and datasets, especially thanks to their free-text inputs. However, they struggle to handle some downstream tasks, such as fine-grained attribute detection and localization. In this paper, we propose a multitask fine-tuning strategy based on a positive/negative prompt formulation to further leverage the capacities of the vision-language foundation models. Using the CLIP architecture as baseline, we show strong improvements on bird fine-grained attribute detection and localization tasks, while also increasing the classification performance on the CUB200-2011 dataset. We provide source code for reproducibility purposes: it is available at https://github.com/FactoDeepLearning/MultitaskVLFM.
• Optical noise characteristics of injection-locked epitaxial quantum dot lasers on silicon
  
  • Chu Qi
  • Zhao Shiyuan
  • Wang Jiawei
  • Sun Yunxu
  • Yao Yong
  • Xu Xiaochuan
  • Grillot Frédéric
  • Duan Jianan
  Optics Express, Optical Society of America - OSA Publishing, 2023, 31 (15), pp.25177-25190. This work theoretically investigates the relative intensity noise (RIN) and spectral linewidth characteristics of epitaxial quantum dot (QD) lasers on silicon subject to optical injection. The results show that the RIN of QD lasers can be reduced by optical injection, hence a reduction of 10 dB is achieved which leads to a RIN as low as −167.5 dB/Hz in the stable injection-locked area. Furthermore, the spectral linewidth of the QD laser can be greatly improved through the optical injection locked scheme. It is reduced from 556.5 kHz to 9 kHz with injection ratio of −60 dB and can be further reduced down to 1.5 Hz with injection ratio of 0 dB. This work provides an effective method for designing low intensity noise and ultra-narrow linewidth QD laser sources for photonics integrated circuits on silicon. (10.1364/oe.492580)
  DOI : 10.1364/oe.492580
• Learn How to Prune Pixels for Multi-View Neural Image-Based Synthesis
  
  • Milovanović Marta
  • Tartaglione Enzo
  • Cagnazzo Marco
  • Henry Félix
  , 2023, pp.158-163. (10.1109/ICMEW59549.2023.00034)
  DOI : 10.1109/ICMEW59549.2023.00034
• All you ever wanted to know about side-channel attacks and protections (and a forthcoming book)
  
  • Cheng Wei
  • Guilley Sylvain
  • Rioul Olivier
  , 2023. Cryptographic chips play fundamental roles in establishing secure systems and secure interconnections among them. However, side-channel attacks utilizing the intrinsic physically observable leakages during the chip running shall reduce or even ruin the security built upon these chips. In this presentation, we give an overview of existing side-channel attack and protection techniques, including those which are involved in certification aspects. We stress that open hardware allows the community to check that protections are sound, which is paramount as they rely on the way they are implemented. This is a perfect illustration where open hardware benefits certification, hence technology adoption and dissemination. A book to be published in Q1 2024 at Springer/Nature will give a mathematical foundation of security guarantees and derivation of optimal attacks on unprotected and protected devices.
• A deep learning method trained on synthetic data for digital breast tomosynthesis reconstruction
  
  • Quillent Arnaud
  • Bismuth Vincent
  • Bloch Isabelle
  • Kervazo Christophe
  • Ladjal Saïd
  , 2023, 227, pp.1813-1825. Digital Breast Tomosynthesis (DBT) is an X-ray imaging modality enabling the reconstruction of 3D volumes of breasts. DBT is mainly used for cancer screening, and is intended to replace conventional mammography in the coming years. However, DBT reconstructions are impeded by several types of artefacts induced by the geometry of the device itself, degrading the image quality and limiting its resolution along the thickness of the compressed breast. In this study, we propose a deep-learning-based pipeline to address the DBT reconstruction problem, focusing on the removal of sparse-view and limited-angle artefacts. Specifically, this procedure is composed of two steps: a classic reconstruction algorithm is first applied on normalised projections, then a deep neural network is tasked with erasing the artefacts present in the obtained volumes. A major difficulty to solve our problem is the lack of real conditions artefact-free data. To overcome this complication, we resort to a new dataset comprised of synthetic breast texture phantoms. We then show that our training method and database strategy are promising to tackle the problem as they improve the informational value of planes orthogonal to the detector, which are not currently used by radiologists due to their poor quality. Eventually, we assess the impact of removing the bias components from the network and using stacks of slices as inputs, with regard to the generalisation ability of our approach on both synthetic and clinical data.
• BiSync: A Bilingual Editor for Synchronized Monolingual Texts
  
  • Crego Josep
  • Xu Jitao
  • Yvon François
  , 2023, pp.369–376. In our globalized world, a growing number of situations arise where people are required to communicate in one or several foreign languages. In the case of written communication, users with a good command of a foreign language may find assistance from computeraided translation (CAT) technologies. These technologies often allow users to access external resources, such as dictionaries, terminologies or bilingual concordancers, thereby interrupting and considerably hindering the writing process. In addition, CAT systems assume that the source sentence is fixed and also restrict the possible changes on the target side. In order to make the writing process smoother, we present BiSync, a bilingual writing assistant that allows users to freely compose text in two languages, while maintaining the two monolingual texts synchronized. We also include additional functionalities, such as the display of alternative prefix translations and paraphrases, which are intended to facilitate the authoring of texts. We detail the model architecture used for synchronization and evaluate the resulting tool, showing that high accuracy can be attained with limited computational resources. The interface and models are publicly available at https://github.com/ jmcrego/BiSync and a demonstration video can be watched on YouTube.
• Disentangled latent representations of images with atomic autoencoders
  
  • Newson Alasdair
  • Traonmilin Yann
  , 2023. We present the atomic autoencoder architecture, which decomposes an image as the sum of elementary parts that are parametrized by simple separate blocks of latent codes. We show that this simple architecture is induced by the definition of a general atomic low-dimensional model of the considered data. We also highlight the fact that the atomic autoencoder achieves disentangled low-dimensional representations under minimal hypotheses. Experiments show that their implementation with deep neural networks is successful at learning disentangled representations on two different examples: images constructed with simple parametric curves and images of filtered off-the-grid spikes.
• Impact of Indoor Distributed Antenna System on RF-EMF Global Exposure
  
  • Mazloum Taghrid
  • Wang Shanshan
  • Wiart Joe
  IEEE Access, IEEE, 2023, 11, pp.70587 - 70597. We aim in the present paper to address the impact of installing indoor distributed antenna system (distAS) on the human exposure to radio-frequency electromagnetic field (RF-EMF). We note that distAS aims to extend coverage and improve wireless communication quality. We performed measurement campaigns in subway stations, where distAS are deployed. The impact of distAS on the exposure is studied by considering two scenarios where distAS are turned either on or off. The electric field strength is measured at different distances to the distAS, for all the frequency bands and operators. The results show that the DL exposure induced by distAS is very low and far away from the standard limits of ICNIRP. (10.1109/ACCESS.2023.3293642)
  DOI : 10.1109/ACCESS.2023.3293642
• Integration of Heterogeneous Components for Co-Simulation
  
  • Jerray Jawher
  • Ameur-Boulifa Rabea
  • Apvrille Ludovic
  , 2023, 1, pp.637-644. Because of their complexity, embedded systems are designed with sub-systems or components taken in charge by different development teams or entities and with different modeling frameworks and simulation tools, depending on the characteristics of each component. Unfortunately, this diversity of tools and semantics makes the integration of these heterogeneous components difficult. Thus, to evaluate their integration before their hardware or software is available, one solution would be to merge them into a common modeling framework. Yet, such a holistic environment supporting many computation and computation semantics seems hard to settle. Another solution we investigate in this paper is to generically link their respective simulation environments in order to keep the strength and semantics of each component environment. The paper presents a method to simulate heterogeneous components of embedded systems in real-time. These components can be described at any abstraction level. Our main contribution is a generic glue that can analyze in real-time the state of different simulation environments and accordingly enforce the correct communication semantics between components. (10.5220/0012134800003538)
  DOI : 10.5220/0012134800003538