Publications

2019

• Seq2VAR: multivariate time series representation with relational neural networks and linear autoregressive model
  
  • Pineau Edouard
  • Razakarivony Sebastien
  • Bonald Thomas
  , 2019. Finding understandable and meaningful feature representation of multivariate time series (MTS) is a difficult task, since information is entangled both in temporal and spatial dimensions. In particular, MTS can be seen as the observation of simultaneous causal interactions between dynamical variables. Standard way to model these interactions is the vector linear autoregression (VAR). The parameters of VAR models can be used as MTS feature representation. Yet, VAR cannot generalize on new samples, hence independent VAR models must be trained to represent different MTS. In this paper, we propose to use the inference capacity of neural networks to overpass this limit. We propose to associate a relational neural network to a VAR generative model to form an encoder-decoder of MTS. The model is denoted Seq2VAR for Sequence-to-VAR. We use recent advances in relational neural network to build our MTS encoder by explicitly modeling interactions between variables of MTS samples. We also propose to leverage reparametrization tricks for binomial sampling in neural networks in order to build a sparse version of Seq2VAR and find back the notion of Granger causality defined in sparse VAR models. We illustrate the interest of our approach through experiments on synthetic datasets.
• Algorithmes gloutons avec la classe
  
  • Zayana Karim
  • Michalak Pierre
  • Beauseigneur Clément
  • Tanoh Hélène
  , 2019. Les algorithmes gloutons offrent une solution pratique, mais pas toujours optimale, à de nombreux problèmes arithmétiques. Nous en donnons ici deux exemples (fractions égyptiennes et algorithme du monnayeur) avant d'entrer dans des considérations plus théoriques.
• Improved Small Molecule Identification through Learning Combinations of Kernel Regression Models
  
  • Brouard Celine
  • Bassé Antoine
  • d'Alché-Buc Florence
  • Rousu Juho
  Metabolites, MDPI, 2019, 9 (8), pp.160. In small molecule identification from tandem mass (MS/MS) spectra, input–output kernel regression (IOKR) currently provides the state-of-the-art combination of fast training and prediction and high identification rates. The IOKR approach can be simply understood as predicting a fingerprint vector from the MS/MS spectrum of the unknown molecule, and solving a pre-image problem to find the molecule with the most similar fingerprint. In this paper, we bring forward the following improvements to the IOKR framework: firstly, we formulate the IOKRreverse model that can be understood as mapping molecular structures into the MS/MS feature space and solving a pre-image problem to find the molecule whose predicted spectrum is the closest to the input MS/MS spectrum. Secondly, we introduce an approach to combine several IOKR and IOKRreverse models computed from different input and output kernels, called IOKRfusion. The method is based on minimizing structured Hinge loss of the combined model using a mini-batch stochastic subgradient optimization. Our experiments show a consistent improvement of top-k accuracy both in positive and negative ionization mode data (10.3390/metabo9080160)
  DOI : 10.3390/metabo9080160
• Addressing Failure and Aging Degradation in MRAM/MeRAM-on-FDSOI Integration
  
  • Cai Hao
  • Wang You
  • Naviner Lirida
  • Liu Xinning
  • Shan Weiwei
  • Yang Jun
  • Zhao Weisheng
  IEEE Transactions on Circuits and Systems I: Regular Papers, IEEE, 2019, 66 (1), pp.239-250. (10.1109/TCSI.2018.2854277)
  DOI : 10.1109/TCSI.2018.2854277
• Improving data collection in complex networks with failure-prone agents via local marking
  
  • Rodrigues Arles
  • Botina Nathaly
  • Gomez Jonatan
  • Diaconescu Ada
  Journal of Intelligent and Fuzzy Systems, IOS Press, 2019.
• Towards Interpretability of Segmentation Networks by analyzing DeepDreams
  
  • Couteaux Vincent
  • Nempont O.
  • Pizaine Guillaume
  • Bloch Isabelle
  , 2019, LCNS 11797, pp.56-63. Interpretability of a neural network can be expressed as the identification of patterns or features to which the network can be either sensitive or indifferent. To this aim, a method inspired by DeepDream is proposed, where the activation of a neuron is maximized by performing gradient ascent on an input image. The method outputs curves that show the evolution of features during the maximization. A controlled experiment show how it enables assess the robustness to a given feature, or by contrast its sensitivity. The method is illustrated on the task of segmenting tumors in liver CT images.
• Automatic Knee Meniscus Tear Detection and Orientation Classification with Mask-RCNN
  
  • Couteaux Vincent
  • Si-Mohamed S.
  • Nempont O.
  • Lefevre T.
  • Popoff A.
  • Pizaine Guillaume
  • Villain N.
  • Bloch Isabelle
  • Cotten A.
  • Boussel L.
  Diagnostic and Interventional Imaging, Elsevier, 2019, 100, pp.235-242.
• Crusts on the Eyelashes
  
  • Ouedraogo Muriel
  • Ventejou Sarah
  • Leducq Sophie
  • Desoubeaux Guillaume
  • Maruani Annabel
  J Pediatr, 2019. A 27-month-old boy was referred to pediatric dermatology for the evaluation of crusts on his eyelashes; they had been present for 2 months and were increasing in number. The crusts had appeared after a trip to Bulgaria, where his family had been sleeping in several different hotels and places. Physical examination showed crusts on both eyelashes and a few crusts on the front part of his scalp. He had slight conjunctivitis and palpebral pruri-tis. The child was in good health and the physical examination was otherwise normal. His weight was 13.5 kg. Findings of a dermoscopy examination revealed that the crusts actually consisted of agglomerates of nits on the prox-imal side of the eyelashes, and several parasites were visible (Figure). There were a few parasites on the front part of his scalp hair also. The hair on his father's chest had several nits. Parasitology examination revealed that the parasites were Phtirus pubis. The child and his parents were given oral ivermectin 400 mg/kg (day 1 and day 7) and topical ophthalmologic rifamycin ointment once a day for 7 days, along with manual removal of the nits and parasites. Complete remission was obtained after 3 weeks of treatment. After interrogation, we concluded that phthiriasis had been transmitted from the father's chest to the child, who used to sleep on the chest of his parents. Phthiriasis palpebrarum is very uncommon in children, especially infants. It can reach the eyelashes but also can extend to the hair and might be complicated by conjunctivitis by an irritative mechanism. 1,2 The clinical diagnosis is confirmed by parasitology examination. The treatment usually consists of mechanical (manual or with tweezers) removal of nits and parasites using petrolatum. In our case, ophthalmologic rifamycin ointment was used to treat the conjunctivitis and facilitated manual removal. We added ivermectin treatment with the dosage recommended for difficult-to-treat head lice, 3,4 which is greater than that for scabies. Phthiriasis in children must always raise the question of the contamination method, which in children is secondary to contact with an infested adult. The origins of this infection need to be carefully discerned, and the question of sexual abuse must be raised. 5 n (10.1016/j.jpeds.2019.02.002)
  DOI : 10.1016/j.jpeds.2019.02.002
• Almost surely constrained convex optimization
  
  • Fercoq Olivier
  • Alacaoglu Ahmet
  • Necoara Ion
  • Cevher Volkan
  , 2019, 97, pp.1910-1919. We propose a stochastic gradient framework for solving stochastic composite convex optimization problems with (possibly) infinite number of linear inclusion constraints that need to be satisfied almost surely. We use smoothing and homotopy techniques to handle constraints without the need for matrix-valued projections. We show for our stochastic gradient algorithm $\mathcal{O}(\log(k)/\sqrt{k})$ convergence rate for general convex objectives and $\mathcal{O}(\log(k)/k)$ convergence rate for restricted strongly convex objectives. These rates are known to be optimal up to logarithmic factors, even without constraints. We demonstrate the performance of our algorithm with numerical experiments on basis pursuit, a hard margin support vector machines and a portfolio optimization and show that our algorithm achieves state-of-the-art practical performance.
• Safe Grid Search with Optimal Complexity
  
  • Ndiaye Eugene
  • Le Tam
  • Fercoq Olivier
  • Salmon Joseph
  • Takeuchi Ichiro
  , 2019, 97, pp.4771-4780. Popular machine learning estimators involve regularization parameters that can be challenging to tune, and standard strategies rely on grid search for this task. In this paper, we revisit the techniques of approximating the regularization path up to predefined tolerance $\epsilon$ in a unified framework and show that its complexity is $O(1/\sqrt[d]{\epsilon})$ for uniformly convex loss of order $d>0$ and $O(1/\sqrt{\epsilon})$ for Generalized Self-Concordant functions. This framework encompasses least-squares but also logistic regression (a case that as far as we know was not handled as precisely by previous works). We leverage our technique to provide refined bounds on the validation error as well as a practical algorithm for hyperparameter tuning. The later has global convergence guarantee when targeting a prescribed accuracy on the validation set. Last but not least, our approach helps relieving the practitioner from the (often neglected) task of selecting a stopping criterion when optimizing over the training set: our method automatically calibrates it based on the targeted accuracy on the validation set.
• Fault Analysis Assisted by Simulation
  
  • Chibani Kais
  • Facon Adrien
  • Guilley Sylvain
  • Marion Damien
  • Mathieu Yves
  • Sauvage Laurent
  • Souissi Youssef
  • Takarabt Sofiane
  , 2019, pp.263-277. Side-channel and fault injection attacks are renown techniques to extract keys from cryptographic devices. Fortunately, leakage reduction and fault detection countermeasures exist and can be implemented right in the source-code. However, source-code level countermeasures might be altered during the compilation process. Design simulation is an effective means to detect such harmful simplifications. This is a well-known methodology to analyze regressions in terms of side-channel leakage. In this chapter, we explain that protections against fault injection attacks are no exception. First of all, we show that vulnerabilities to those attacks can be easily detected by simulation methods. Second, we highlight that simulation techniques are also highly efficient in detecting logic simplifications which destroy (fully or partly) the countermeasures. Thus, the simulation-based methodology we present in this chapter shows that it is possible to decide quickly which compilation options are safe and which ones are detrimental to the security. (10.1007/978-3-030-11333-9_12)
  DOI : 10.1007/978-3-030-11333-9_12
• Learning-based tone mapping operator for efficient image matching
  
  • Rana Aakanksha A
  • Valenzise Giuseppe
  • Dufaux Frédéric
  IEEE Transactions on Multimedia, Institute of Electrical and Electronics Engineers, 2019, 21 (1), pp.256-268. In this paper, we propose a new framework to optimally tone map the high dynamic range (HDR) content for image matching under drastic illumination variations. Since tone mapping operators (TMO) have traditionally been used for displaying HDR scenes, their design is suboptimal when used for computer vision tasks such as image matching. We address this sub-optimality by proposing a two-step framework, consisting of: a) a luminance-invariant guidance model based on a Support Vector Regressor (SVR) to optimally adapt the tone mapping function for image matching; and b) an energy maximization model to generate appropriate training samples for learning the SVR. At each step, we collectively address both stages of keypoint detection and descriptor extraction in the feature matching framework. By locally altering the intrinsic characteristics of the tone mapping function, the learned guid- ance model facilitates the extraction of local invariant features in the presence of illumination variations. We demonstrate that the proposed TMO significantly outperforms perceptually-driven state-of-the-art TMOs on a dataset of HDR scenes characterized by challenging lighting variations, such as day/night transitions. (10.1109/TMM.2018.2839885)
  DOI : 10.1109/TMM.2018.2839885
• Ontology-mediated query answering over temporal and inconsistent data
  
  • Bourgaux Camille
  • Koopmann Patrick
  • Turhan Anni-Yasmin
  Semantic Web – Interoperability, Usability, Applicability, IOS Press, 2019. Stream-based reasoning systems process data stemming from different sources that are received over time. In this kind of applications, reasoning needs to cope with the temporal dimension and should be resilient against inconsistencies in the data. Motivated by such settings, this paper addresses the problem of handling inconsistent data in a temporal version of ontology-mediated query answering. We consider a recently proposed temporal query language that combines conjunctive queries with operators of propositional linear temporal logic (LTL), and consider these under three inconsistency-tolerant semantics that have been introduced for querying inconsistent description logic knowledge bases. We investigate their complexity for temporal EL ⊥ and DL-LiteR knowledge bases. In particular, we consider two different cases, depending on the presence of negations in the query. Furthermore, we complete the complexity picture for the consistent case. We also provide two approaches toward practical algorithms for inconsistency-tolerant temporal query answering. (10.3233/SW-180337)
  DOI : 10.3233/SW-180337
• Complexity of Unique (Optimal) Solutions in Graphs: Vertex Cover and Domination
  
  • Hudry Olivier
  • Lobstein Antoine
  Journal of Combinatorial Mathematics and Combinatorial Computing, Charles Babbage Research Centre, 2019, 110, pp.217-240. We study the complexity of four decision problems dealing with the uniqueness of a solution in a graph: "Uniqueness of a Vertex Cover with bounded size"(U-VC) and "Uniqueness of an Optimal Vertex Cover"(U-OVC), and for any fixed integer r ≥ 1, "Uniqueness of an r-Dominating Code with bounded size" (U-DCr) and "Uniqueness of an Optimal r-Dominating Code" (U-ODCr). In particular, we give a polynomial reduction from "Unique Satisfiability of a Boolean formula" (U-SAT
• Frequency comb dynamics of a 13 μm hybrid-silicon quantum dot semiconductor laser with optical injection
  
  • Beausoleil Raymond G
  • Grillot Frédéric
  • Dong Bozhang
  • Huang Heming
  • Duan Jianan
  • Kurczveil Geza
  • Liang Di I
  Optics Letters, Optical Society of America - OSA Publishing, 2019, 44 (23), pp.5755. This work reports on the influence of bias voltage applied on a saturable absorber (SA) on a subthreshold linewidth enhancement factor (LEF) in hybrid-silicon quantum dot optical frequency comb lasers. Results show that the reverse bias voltage on SA contributes to enlarge the LEF and improve the comb dynamics. Optical injection is also found to be able to improve the comb spectrum in terms of 3 dB bandwidth and its flatness. Such novel findings are promising for the development of high-speed dense wavelength-division multiplexing photonic integrated circuits in optical interconnects and datacom applications. (10.1364/OL.44.005755)
  DOI : 10.1364/OL.44.005755
• •Intelligent Systems for Crisis Management: Gi4DM 2018
  
  • Altan Orhan
  • Chandra Madhu
  • Sunar F
  • Tanzi Tullio
  Lecture Notes in Geoinformation and Cartography, Springer International Publishing, 2019. In the past several years, there have been significant technological advances in the field of crisis response. However, many aspects concerning the efficient collection and integration of geo-information, applied semantics and situation awareness for disaster management remain open. Improving crisis response systems and making them intelligent requires extensive collaboration between emergency responders, disaster managers, system designers and researchers alike. To facilitate this process, the Gi4DM (GeoInformation for Disaster Management) conferences have been held regularly since 2005. The events are coordinated by the Joint Board of Geospatial Information Societies (JB GIS) and ICSU GeoUnions. This book presents the outcomes of the Gi4DM 2018 conference, which was organised by the ISPRS-URSI Joint Working Group ICWG III/IVa: Disaster Assessment, Monitoring and Management and held in Istanbul, Turkey on 18-21 March 2018. It includes 12 scientific papers focusing on the intelligent use of geo-information, semantics and situation awareness. (10.1007/978-3-030-05330-72019)
  DOI : 10.1007/978-3-030-05330-72019
• A statistical detection mechanism for node misbehaviours in wireless mesh networks
  
  • Khatoun Rida
  • Begriche Youcef
  • Khoukhi Lyes
  IJAHUC - International Journal of Ad Hoc and Ubiquitous Computing, Inderscience, 2019, 31 (1), pp.23. (10.1504/IJAHUC.2019.099637)
  DOI : 10.1504/IJAHUC.2019.099637
• Compressed sensing-enabled phase-sensitive swept-source optical coherence tomography
  
  • Ling Yuye
  • Meiniel William
  • Singh-Moon Rajinder
  • Angelini Elsa D.
  • Olivo-Marin J.-C.
  • Hendon Christine P.
  Optics Express, Optical Society of America - OSA Publishing, 2019, 27 (2), pp.855-871.
• Evolving, Growing, and Gardening Cyber-physical Systems
  
  • Stepney Susan
  • Diaconescu Ada
  • Doursat René
  • Giavitto Jean-Louis
  • Miller Julian
  • Spicher Antoine
  , 2019.
• Analysis of CEDBT and CESM performance using a realistic X-ray simulation platform
  
  • Sanchez de La Rosa Ruben
  • Carton A.-K.
  • Milioni de Carvalho P.
  • Bloch Isabelle
  • Muller S.
  , 2019, pp.1070-1073. Contrast Enhanced Spectral Mammography (CESM) and Contrast Enhanced Digital Breast Tomosynthesis (CEDBT) are multi-energy X-ray imaging techniques involving the injection of a vascular contrast agent. Both techniques provide information on hypervascularization of lesions through contrast uptake. CESM has proved to deliver a better diagnosis of breast cancer than diagnostic mammography. CEDBT is a promising technique which provides 3D information on the contrast uptake distribution. In this paper, new steps in the image acquisition process of a previously presented image acquisition simulation platform are described, including models of scatter, image lag and electronic noise. Using this simulation platform, 290 CESM and CEDBT images were generated. A human observer experiment was then performed to compare lesion detectability and characterization. The results indicate a similar detectability and an improved characterization of shape and contrast enhancement distribution using CEDBT.
• Segmentation et caractérisation des bifurcations artérielles rétiniennes dans des images 2D d’optique adaptative
  
  • Trimeche Iyed
  • Rossant Florence
  • Bloch Isabelle
  • Pâques M.
  , 2019. Nous pr ́esentons une m ́ethode de segmentation des art`eres r ́etiniennes dans des images de fond d’oeil de haute r ́esolution, acquises en optique adaptative. Nous e ́ tendons notre approche pr ́ec ́edente de traitement des branches des vaisseaux r ́etiniens a ` la segmentation des bifur- cations, ce qui nous permet d’analyser l’int ́egralit ́e de l’arbre vasculaire. Diff ́erents biomarqueurs caract ́erisant le flux sanguin sont extraits de l’estimation des diam`etres des branches aux bifurcations. Les r ́esultats exp ́erimentaux montrent que la pr ́ecision de notre approche se situe dans la plage de variabilit ́e intra- et inter-utilisateurs, ce qui nous a permis de r ́ealiser une e ́ tude pr ́eliminaire sur les biomarqueurs extraits.
• Suivi des glaciers de montagne par imagerie radar satellitaire
  
  • Fallourd Renaud
  • Dehecq Amaury
  • Jauvin Matthias
  • Yan Yajing
  • Vasile Gabriel
  • Gay Michel
  • Trouvé Emmanuel
  • Nicolas Jean Marie
  Revue Française de Photogrammétrie et de Télédétection, Société Française de Photogrammétrie et de Télédétection, 2019 (219-220), pp.91-105. Cet article présente un ensemble de résultats obtenus par télédétection radar satellitaire sur le site test Chamonix-Mont Blanc. L'objectif est d'illustrer le potentiel et les limitations de l'imagerie SAR (Synthetic Aperture Radar) pour l'observation des glaciers de montagne en zone tempérée. Après avoir rappelé certaines spécificités de ces glaciers qui conditionnent l'exploitation des données SAR, trois types de mesures sont étudiés : i/ le calcul de champs de déformation par interférométrie différentielle (D-InSAR) avec des données ERS Tandem dont les couplesà un jour permettaient de mettre en œuvre cette technique en dehors de la période estivale ; ii/ le calcul de champs de déformation bidimensionnels par corrélation d'amplitude avec des données haute résolution TerraSAR-X (TSX) et la reconstruction du déplacement tridimensionnelà partir de couples acquis sur des orbites ascendantes et descendantes ; iii/ le calcul de la topographieà partir de couples TanDEM-X (TDX) en évaluant les incertitudes liéesà la pénétration radar dans la neige et la glace. Enfin, nous illustrons les premiers résultats obtenus avec les données des satellites Sentinel-1 A/B, de plus faible résolution spatiale mais dont la répétitivitéà 6 jours et la gratuité ouvrent de nouvelles perspectives. Ces travaux montrentà la fois le fort potentiel de l'imagerie SAR pour observer la dynamique et les variations de volume des glaciers tout en soulignant les facteurs favorables ou limitant pour une exploitation régulière des données issues des satellites radar lancés depuis les années 90. (10.52638/rfpt.2019.471)
  DOI : 10.52638/rfpt.2019.471