Publications

2018

• Feedback increases the capacity of queues
  
  • Tchamkerten Aslan
  • Aptel Laure
  , 2018, pp.1116-1120. In their “Bits Through Queues” paper, Anantharam and Verdu showed that, under FIFO service policy, feedback does not increase the capacity of a queue when the service time is exponentially distributed. Whether this negative and surprising result-since the channel has memory-holds for other combinations of service policies and service times ever since has remained an open question. This paper first provides a sufficient condition on the service time distribution for feedback to increase capacity under First-In-First-Out service policy. Underlying this condition is a notion of weak feedback wherein instead of the queue departure times the transmitter is informed about the instants when packets start to be served. Service times that satisfy this condition include a uniformly distributed service time for the continuous-time model and a binary service time for the discrete-time model. Second, a sufficient condition is given under which feedback does not increase capacity. This condition is satisfied, for instance, by queues with Last-Come- First-Served service policies and bounded service times. (10.1109/ISIT.2018.8437726)
  DOI : 10.1109/ISIT.2018.8437726
• Confusing information: How confusion improves side-channel analysis for monobit leakages
  
  • de Chérisey Eloi
  • Guilley Sylvain
  • Rioul Olivier
  , 2018. In this presentation, we will recall the notions of confusion and show that, in the case of a monobit leakage, MIA and CPA are similar. To do so, we recall the link between CPA and the confusion coecient, and then, explore by Taylor expansions how we can link this confusion coecient to MIA. We show that the distinguishers corresponding to both CPA and MIA are proportional when the sensitive variable is binary. We apply these results to the case of a monobit leakage with AES substitution box and without substitution box to show the impact of this confusion coecient, and how this impact can be measured. Finally, we open the discussion to extend these results to non-binary distributions.
• Distributed Hypothesis Testing with Concurrent Detections
  
  • Escamilla Pierre
  • Wigger Michèle
  • Zaidi Abdellatif
  , 2018, pp.166-170. A detection system with a single sensor and $\mathsf{K}$ detectors is considered, where each of the terminals observes a memoryless source sequence and the sensor sends a common message to all the detectors. The communication of this message is assumed error-free but rate-limited. The joint probability mass function (pmf) of the source sequences observed at the terminals depends on an $\mathsf{M}$-ary hypothesis $(\mathsf{M} \geq \mathsf{K})$, and the goal of the communication is that each detector can guess the underlying hypothesis. Each detector $k$ aims to maximize the error exponent under hypothesis $k$, while ensuring a small probability of error under all other hypotheses. This paper presents an achievable exponents region for the case of positive communication rate, and characterizes the optimal exponents region for the case of zero communication rate. All results extend also to a composite hypothesis testing scenario. (10.1109/ISIT.2018.8437906)
  DOI : 10.1109/ISIT.2018.8437906
• EBP-GEXIT charts over the binary-input AWGN channel for generalized and doubly-generalized LDPC codes
  
  • Yardi Arti D
  • Andriyanova Iryna
  • Poulliat Charly
  , 2018. This work proposes a tractable evaluation of the maximum a posteriori (MAP) threshold of sparse-graph ensembles, by using an approximation for the extended belief propagation generalized extrinsic information transfer (EBP-GEXIT) function, first proposed by Measson et al. The approximation allows to find a MAP threshold in such numerically involved cases as the binary-input additive white Gaussian noise (AWGN) channel, graph ensembles with general component codes and/or irregularities. The paper contains examples of estimations of the MAP thresholds in the case of irregular low-density parity-check (LDPC), generalized LDPC, and doubly generalized LDPC codes ensembles. Our estimations are confirmed by numerical simulations.
• Parametric versus nonparametric: the fitness coefficient
  
  • Mazo Gildas
  • Portier François
  , 2018. The fitness coefficient, introduced in this paper, results from a competition between parametric and nonparametric density estimators within the likelihood of the data. As illustrated on several real datasets, the fitness coefficient generally agrees with p-values but is easier to compute and interpret. Namely, the fitness coefficient can be interpreted as the proportion of data coming from the parametric model. Moreover, the fitness coefficient can be used to build a semiparamteric compromise which improves inference over the parametric and nonparametric approaches. From a theoretical perspective, the fitness coefficient is shown to converge in probability to one if the model is true and to zero if the model is false. From a practical perspective, the utility of the fitness coefficient is illustrated on real and simulated datasets.
• Domain adaptation with optimal transport improves EEG sleep stage classifiers
  
  • Chambon Stanislas
  • Galtier Mathieu N.
  • Gramfort Alexandre
  , 2018. —Low sample size and the absence of labels on certain data limits the performances of predictive algorithms. To overcome this problem, it is sometimes possible to learn a model on a large labeled auxiliary dataset. Yet, this assumes that the two datasets exhibit similar statistical properties which is rarely the case in practice: there is a discrepancy between the large dataset, called the source, and the dataset of interest, called the target. Improving the prediction performance on the target domain by reducing the distribution discrepancy, between the source and the target domains, is known as Domain Adaptation (DA). Presently, Optimal transport DA (OTDA) methods yield state-of-the-art performances on several DA problems. In this paper, we consider the problem of sleep stage classification, and use OTDA to improve the performances of a convolutional neural network. We use features learnt from the electroencephalogram (EEG) and the electrooculogram (EOG) signals. Our results demonstrate that the method significantly improves the network predictions on the target data.
• Augmenting DER hosting capacity of distribution grids through local energy markets and dynamic phase switching
  
  • Horta José
  • Kofman Daniel
  • Menga David
  • Caujolle Mathieu
  , 2018, 18. The limited capacity of distribution grids for hosting renewable generation is one of the main challenges towards the energy transition. Local energy markets, enabling direct exchange of energy between prosumers, help to integrate the growing number of residential photovoltaic panels by scheduling flexible demand for balancing renewable energy locally. Nevertheless, existing scheduling mechanisms do not take into account the phases to which households are connected, increasing network unbalance and favoring bigger voltage rises/drops and higher losses. In this paper, we reduce network unbalance by leveraging market transactions information to dynamically allocate houses to phases using solid state switches. We propose cost effective mechanisms for the selection of households to switch and for their optimal allocation to phases. Using load flow analysis we show that only 6% of houses in our case studies need to be equipped with dynamic switches to counteract the negative impact of local energy markets while maintaining all the benefits. Combining local energy markets and dynamic phase switching we improve both overall load balancing and network unbalance, effectively augmenting DER hosting capacity of distribution grids. (10.1145/3208903.3208937)
  DOI : 10.1145/3208903.3208937
• When Can We Answer Queries Using Result-Bounded Data Interfaces?
  
  • Amarilli Antoine
  • Benedikt Michael
  , 2018. We consider answering queries where the underlying data is available only over limited interfaces which provide lookup access to the tuples matching a given binding, but possibly restricting the number of output tuples returned. Interfaces imposing such "result bounds" are common in accessing data via the web. Given a query over a set of relations as well as some integrity constraints that relate the queried relations to the data sources, we examine the problem of deciding if the query is answerable over the interfaces; that is, whether there exists a plan that returns all answers to the query, assuming the source data satisfies the integrity constraints. The first component of our analysis of answerability is a reduction to a query containment problem with constraints. The second component is a set of "schema simplification" theorems capturing limitations on how interfaces with result bounds can be useful to obtain complete answers to queries. These results also help to show decidability for the containment problem that captures answerability, for many classes of constraints. The final component in our analysis of answerability is a "linearization" method, showing that query containment with certain guarded dependencies -- including those that emerge from answerability problems -- can be reduced to query containment for a well-behaved class of linear dependencies. Putting these components together, we get a detailed picture of how to check answerability over result-bounded services.
• Caractérisation d'excitation de caisses de guitare : vers l'identification des efforts bloqués au chevalet
  
  • Dujourdy Hugo
  • Ablitzer Frédéric
  • Cabaret Jérémy
  • David Bertrand
  • Gautier François
  , 2018. En acoustique musicale et plus largement au niveau industriel, la transmission des puissances vibratoires par une structure source vers une structure réceptrice dépend des caractéristiques dynamiques de ces deux structures. Elle peut être prédite par la mesure des forces bloquées, dont l’identification est un enjeu majeur pour la caractérisation des systèmes couplés. Sur un instrument à cordes, les forces bloquées au chevalet correspondent aux efforts transmis par les cordes si celles-ci étaient montées sur un banc rigide : leur mesure en fonction du temps permettra de comparer différents instruments pour un même geste musical ou même d’alimenter des modèles de synthèse sonore. Cette mesure ne peut être effectuée qu’en situation de jeu, et la mesure in-situ et indirecte des forces bloquées est bien décrite dans la littérature. Nous développons une méthode inverse pour l'identification des paramètres du jeu au niveau du chevalet d’instruments à cordes pincées ou frottées mesurées in-situ et dans le domaine temporel. Quelques expériences permettant de démontrer la faisabilité de la méthode seront présentés.
• 10-100 Gbit/s mixed-line-rate experimental demonstration
  
  • Chouman Hussein
  • Lourdiane Mounia
  • Tremblay Christine
  • Adhikari Bibechana
  • Ware Cédric
  • Lepers Catherine
  , 2018 (OP-COMM-1-3-3). In this paper we analyze the feasibility of mixed-line rate transmission of 10 and 100 Gbit/s links for metro and long haul optical network. We compare theoretical models, simulation and experimental results for different transmission scenarios
• Mathematical Morphology on Tensor Images for Fiber Enhancement
  
  • Blusseau Samy
  • Bloch Isabelle
  • Gousseau Y.
  • Velasco-Forero S
  • Angulo Jesus
  , 2018. We introduce morphological methods to analyse images of positive semi-definite matrices (ellipsoids), such as diffusion or structure tensors. Morphological operators cannot be directly defined on these objects because they lack a complete lattice structure. Therefore, we extend recent developments in morphological signal processing on graphs. Graphs are constructed based on all the available information, whereas a scalar signal (anisotropy) is processed on these graphs, successfully enhancing structures such as vessels and fibers.
• Form Filling based on Constraint Solving
  
  • Spencer Ben
  • Benedikt Michael
  • Senellart Pierre
  , 2018, 10845. We describe a system for analyzing form-based websites to discover sequences of actions and values that result in a valid form submission. Rather than looking at the text or DOM structure of the form, our method is driven by solving constraints involving the underlying client-side JavaScript code. In order to deal with the complexity of client-side code, we adapt a method from program analysis and testing, concolic testing, which mixes concrete code execution, symbolic code tracing, and constraint solving to find values that lead to new code paths. While concolic testing is commonly used for detecting bugs in stand-alone code with developer support, we show how it can be applied to the very different problem of filling Web forms. We evaluate our system on a benchmark of both real and synthetic Web forms. (10.1007/978-3-319-91662-0_20)
  DOI : 10.1007/978-3-319-91662-0_20
• Focused Crawling through Reinforcement Learning
  
  • Han Miyoung
  • Wuillemin Pierre-Henri
  • Senellart Pierre
  , 2018, 10845, pp.261-278. Focused crawling aims at collecting as many Web pages relevant to a target topic as possible while avoiding irrelevant pages, reflecting limited resources available to a Web crawler. We improve on the efficiency of focused crawling by proposing an approach based on reinforcement learning. Our algorithm evaluates hyperlinks most profitable to follow over the long run, and selects the most promising link based on this estimation. To properly model the crawling environment as a Markov decision process, we propose new representations of states and actions considering both content information and the link structure. The size of the state-action space is reduced by a generalization process. Based on this generalization, we use a linear-function approximation to update value functions. We investigate the trade-off between synchronous and asynchronous methods. In experiments, we compare the performance of a crawling task with and without learning; crawlers based on reinforcement learning show better performance for various target topics. (10.1007/978-3-319-91662-0_20)
  DOI : 10.1007/978-3-319-91662-0_20
• A common data fusion framework for space robotics: architecture and data fusion methods
  
  • Dominguez Raúl
  • Michalec Romain
  • Oumer Nassir
  • Souvannavong Fabrice
  • Post Mark
  • Govindaraj Shashank
  • Fabisch Alexander
  • Wehbe Bilal
  • Gancet Jérémi
  • Bianco Alessandro
  • Lacroix Simon
  • de Maio Andrea
  • Bissonnette Vincent
  • Smíšek Michal
  • Yan Xiu
  , 2018. Data fusion algorithms make it possible to combine data from different sensors into symbolic representations such as environment maps, object models, and position estimates. The software community in space robotics lacks a comprehensive software framework to fuse and contextually store data from multiple sensors while also making it easier to develop, evaluate, and compare algorithms. The InFuse consortium, six partners in the industrial and academic space sector working under the supervision of a Program Support Activity (PSA) consisting of representatives from ESA, ASI, CDTI, CNES, DLR, UKSA, is developing such a framework, complete with a set of data fusion implementations based on state-of-the-art perception, localization and mapping algorithms, and performance metrics to evaluate them. This paper describes the architecture of this Common Data Fusion Framework and overviews the data fusion methods that it will provide for tasks such as localisation, mapping, environment reconstruction, object detection and tracking.
• Speckle reduction in PolSAR by multi-channel variance stabilization and Gaussian denoising: MuLoG
  
  • Deledalle Charles-Alban
  • Denis Loïc
  • Tupin Florence
  • Lobry Sylvain
  , 2018.
• SARA
  
  • Monteuuis Jean-Philippe
  • Boudguiga Aymen
  • Zhang Jun
  • Labiod Houda
  • Servel Alain
  • Urien Pascal
  , 2018, pp.3-14. (10.1145/3198458.3198465)
  DOI : 10.1145/3198458.3198465
• De l'analyse d'opinions à la détection des problèmes d'interactions humain-machine : application à la gestion de la relation client
  
  • Poltavchenko Irina
  , 2018. Motivée par le gain en popularité des chatbots prenant le rôle de conseillers sur les sites Web des entreprises, cette thèse s'attaque au problème de la détection des problèmes d’interaction entre un conseiller virtuel et ses utilisateurs sous l'angle de l'analyse des opinions et des émotions dans les textes. Cette thèse s’est déroulée dans le cadre d’une application concrète pour l’entreprise EDF et s'est appuyée sur le corpus du chatbot d'EDF. Ce corpus regroupe des expressions spontanées et riches, collectées dans les conditions écologiques (parfois appelées « in-the-wild »), difficiles à analyser de façon automatique, et encore peu étudiées. Nous proposons une typologie des problèmes d’interaction et faisons annoter une partie du corpus selon cette typologie, annotation dont une partie servira à l’évaluation du système. Le système de Détection Automatique des Problèmes d’Interaction (DAPI) développé lors de cette thèse est un système hybride qui allie l’approche symbolique et l’apprentissage non supervisé de représentation sémantique par plongements lexicaux (word embeddings). Le système DAPI a pour vocation d'être directement connecté au chatbot et de détecter des problèmes d’interaction en ligne, dès la réception d’un énoncé utilisateur. L'originalité de la méthode proposée repose sur : i) la prise en compte de l'historique du dialogue; ii) la modélisation des problèmes d’interaction en tant qu'expressions des opinions et des phénomènes reliés aux opinions spontanées de l'utilisateur vis-à-vis de l'interaction; iii) l'intégration des spécificités du langage web et « in-the-wild » comme des indices linguistiques pour les règles linguistiques; iv) recours aux plongements lexicaux de mots (word2vec) appris sur le grand corpus du chatbot non étiqueté afin de modéliser des similarités sémantiques. Les résultats obtenus sont très encourageants compte tenu de la complexité des données : F-score = 74,3%.
• C-ITS use cases: study, extension and classification methodology
  
  • Haidar Farah
  • Kaiser Arnaud
  • Lonc Brigitte
  • Urien Pascal
  • Denis Richard
  , 2018. —In the near future vehicles will be connected and able to communicate with their environment. Such technologies – commonly called Cooperative Intelligent Transportation Systems (C-ITS) – aim at improving road safety, traffic efficiency and drivers comfort. To this end the C-ITS community has proposed many different use cases. In this paper, we start by making an inventory of C-ITS use cases. We then extend this list by proposing new use cases mostly related to security and privacy aspects. Finally we propose a classification methodology based on K-means algorithm to classify the use cases according to criteria we defined. We apply the proposed methodology on our use cases list using security and technical criteria. The obtained results enable to extract a subset of representative use cases from the initial list. Such subset can then be used to apply any process/method (e.g. risk analysis) on it.
• Feasibility Study of Misbehavior Detection Mechanisms in Cooperative Intelligent Transport Systems (C-ITS)
  
  • Kamel Joseph
  • Kaiser Arnaud
  • Jemaa Ines Ben
  • Cincilla Pierpaolo
  • Urien Pascal
  , 2018. —Cooperative Intelligent Transport Systems (C–ITS) is an emerging technology that aims at improving road safety, traffic efficiency and drivers experience. To this end, vehicles cooperate with each others and the infrastructure by exchanging Vehicle–to–X communication (V2X) messages. In such communicating systems message authentication and privacy are of paramount importance. The commonly adopted solution to cope with these issues relies on the use of a Public Key Infrastructure (PKI) that provides digital certificates to entities of the system. Even if the use of pseudonym certificates mitigate the privacy issues, the PKI cannot address all cyber threats. That is why we need a mechanism that enable each entity of the system to detect and report misbehaving neighbors. In this paper, we provide a state-of-the-art of misbehavior detection methods. We then discuss their feasibility with respect to current standards and law compliance as well as hardware/software requirements.
• Efficient Exact and Approximate Algorithms for Computing Betweenness Centrality in Directed Graphs
  
  • Chehreghani Mostafa Haghir
  • Bifet Albert
  • Abdessalem Talel
  , 2018, 10939, pp.752-764. In this paper, first given a directed network G and a vertex r \in V(G) , we propose a new exact algorithm to compute betweenness score of r. Our algorithm pre-computes a set {RF}(r), which is used to prune a huge amount of computations that do not contribute in the betweenness score of r. Then, for the cases where {RF}(r) is large, we present a randomized algorithm that samples from {RF}(r) and performs computations for only the sampled elements. We show that this algorithm provides (\epsilon ,\delta ) an -approximation of the betweenness score of r. Finally, we empirically evaluate our algorithms and show that they significantly outperform the most efficient existing algorithms, in terms of both running time and accuracy. Our experiments also show that our proposed algorithms can effectively compute betweenness scores of all vertices in a set of vertices. (10.1007/978-3-319-93040-4_59)
  DOI : 10.1007/978-3-319-93040-4_59
• Demoing Platypus – A Multilingual Question Answering Platform for Wikidata
  
  • Pellissier Tanon Thomas
  • Dias de Assuncao Marcos
  • Caron Eddy
  • Suchanek Fabian M.
  , 2018, 11155, pp.111-116. In this paper we present Platypus, a natural language question answering system on Wikidata. Our platform can answer complex queries in several languages, using hybrid grammatical and template based techniques. Our demo allows users either to select sample questions , or formulate their own – in any of the 3 languages that we currently support. A user can also try out our Twitter bot, which replies to any tweet that is sent to its account. (10.1007/978-3-319-98192-5_21)
  DOI : 10.1007/978-3-319-98192-5_21
• Linear UCB for Online SON Management
  
  • Daher Tony
  • Jemaa Sana Ben
  • Decreusefond Laurent
  , 2018. —Policy Based SON Management (PBSM) is the process of orchestrating the deployed Self-Organized Network (SON) functions, so that the network responds as a whole to the operator objectives. This process is based on the configuration of the SON functions in order to steer their actions in the network towards certain operator objectives. The PBSM ensures an automated translation of these objectives into configurations of the SON functions. An approach has been recently proposed to empower the PBSM with cognitive capabilities (C-PBSM), using a Multi-Armed Bandit algorithm, namely the UCB1. The C-PBSM learns the optimal SON configurations based on network feedback. In this paper we propose an alternative approach, based on the LinUCB algorithm, that is able to learn the optimal SON configuration much faster than the previous approach. The speed of convergence is a critical factor that has to be thoroughly considered in the deployment of online learning processes. Results are shown using an LTE-A simulator that considers real-like network topology and parameters, and accurate ray tracing based propagation.
• Positioning data-rate trade-off in mm-wave small cells and service differentiation for 5g networks
  
  • Ghatak G.
  • Koirala R.
  • de Domenico A.
  • Denis B.
  • Dardari D.
  • Uguen Bernard
  , 2018, 2018-June, pp.1-5. We analyze a millimeter wave network, deployed along the streets of a city, in terms of positioning and downlink data-rate performance, respectively. First, we present a transmission scheme where the base stations provide jointly positioning and data-communication functionalities. Accordingly, we study the trade- off between the localization and the data rate performance based on theoretical bounds. Then, we obtain an upper bound on the probability of beam misalignment based on the derived localization error bound. Finally, we prescribe the network operator a scheme to select the beamwidth and the power splitting factor between the localization and communication functions to address different quality of service requirements, while limiting cellular outage. © 2018 IEEE. (10.1109/VTCSpring.2018.8417791)
  DOI : 10.1109/VTCSpring.2018.8417791
• Concomitant sorting of circulating tumor cell subpopulations in cancer patients’ blood samples through the multiplexed immunosensor combination design
  
  • Tran Phuong
  • Tendero Yohann
  • Goldwasser Francois
  • Ropert Stanislas
  • Coriat Romain
  International Biology Review, 2018, 2 (2), pp.1-15.
• WebLinux: a scalable in-browser and client-side Linux and IDE
  
  • Sharrock Rémi
  • Angrave Lawrence
  • Hamonic Ella
  , 2018 (Article No. 45). “WebLinux” is a web app tool providing a standard Linux OS and an IDE in the browser, including a terminal, a code editor and a file browser. It provides a client-side and offline Linux OS environment based on a Javascript emulated processor. By avoiding the use of a Virtual Machine or any Linux server, Weblinux enables learners to directly start experimenting with the Linux OS without installing any software. The tool is entirely client-side which makes it extremely scalable and easy to deploy within a large community of online learners. (10.1145/3231644.3231703)
  DOI : 10.1145/3231644.3231703