Search

International audience; Deep learning architectures, albeit successful in most computer vision tasks, were designed for data with an underlying Euclidean structure, which is not usually fulfilled since pre-processed data may lie on a non-linear space. In this paper, we propose a geometric deep learning approach using rigid and non-rigid transformations, named KShapenet, for 2D and 3D landmark-based human motion analysis. Landmark configuration sequences are first modeled as trajectories on Kendall's shape space and then mapped to a linear tangent space. The resulting structured data are then input to a deep learning architecture, which includes a layer that optimizes over rigid and non-rigid transformations of landmark configurations, followed by a CNN-LSTM network. We apply KShapenet to 3D human landmark sequences for action and gait recognition, and 2D facial landmark sequences for expression recognition, and demonstrate the competitiveness of the proposed approach with respect to state-of-the-art.

Efficient sampling from a high-dimensional Gaussian distribution is an old but high-stake issue. Vanilla Cholesky samplers imply a computational cost and memory requirements which can rapidly become prohibitive in high dimension. To tackle these issues, multiple methods have been proposed from different communities ranging from iterative numerical linear algebra to Markov chain Monte Carlo (MCMC) approaches. Surprisingly, no complete review and comparison of these methods have been conducted. This paper aims at reviewing all these approaches by pointing out their differences, close relations, benefits and limitations. In addition to this state of the art, this paper proposes a unifying Gaussian simulation framework by deriving a stochastic counterpart of the celebrated proximal point algorithm in optimization. This framework offers a novel and unifying revisit of most of the existing MCMC approaches while extending them. Guidelines to choose the appropriate Gaussian simulation method for a given sampling problem in high dimension are proposed and illustrated with numerical examples., 53 pages, 11 figures

International audience; A photon structure is advanced based on the experimental evidence and the vector potential quantization at a single photon level. It is shown that the photon is neither a point particle nor an infinite wave but behaves rather like a local "wave-corpuscle" extended over a wavelength, occupying a minimum quantization volume and guided by a non-local vector potential real wave function. The quantized vector potential oscillates over a wavelength with circular left or right polarization giving birth to orthogonal magnetic and electric fields whose amplitudes are proportional to the square of the frequency. The energy $\hbar$$\omega$ and momentum $\hbar$$\bar{k}$ are carried by the local wavecorpuscle guided by the non-local vector potential wave function suitably normalized.

International audience

International audience; Deprecations are a common way to indicate that a given feature or API will not be available in subsequent versions of a library or framework. While raising deprecation warnings lets developers of libraries evolve their APIs, the developers of client applications often have to manually rewrite their applications to adapt to the deprecation (removal, new APIs.. .). Some may use static analysis tools to support the rewriting. However, dynamically-typed languages or the use of reflective features often produce incorrect rewrite candidates. This is a costly activity that can lead to bug introductions. In this article, we present a method deprecation approach and a tool called DEPREWRITER that can automatically rewrite the callers of deprecated methods during program execution. Clients of a deprecated API execute their program and associated tests, and DEPREWRITER dynamically rewrites the source code of methods that called a deprecated API to use the new API. The implementation of DEPREWRITER is based on dynamic program transformation: when a deprecated method is executed, a program transformation engine rewrites and recompiles the caller's code before continuing the execution. The approach presented in this article has been developed by the Pharo consortium. Since 2016, DEPREWRITER is used in production in multiple distributions of the Pharo programming language: Pharo 6, 7, 8, and 9 alpha. This article presents and validates this approach. The validation is done in two steps: first with an analysis of deprecations available in Pharo 8 and second with an open survey of software developers about DEPREWRITER. We studied 367 Pharo 8 deprecations, among which we analyzed the 218 rewriting deprecations that use transformation rules. We identified the validity conditions and reported defects to the community. We also proposed 33 transformation rules to be added to the non-rewriting deprecations. Both contributions were accepted into Pharo 9 alpha. We classified the rules and identified possible points of improvement. In addition, we performed a user survey and collected information from 46 software developers: some of them used existing DEPREWRITER' rules and executed them on their code, others used DEPREWRITER to create rewriting deprecations, and finally, some were not aware of DEPREWRITER. 28 of 46 developers (60%) reported that the rewriting deprecations helped them, while 10 stated the inverse and 8 were uncertain. After discussing the current implementation, we sketch possible implementations for other languages than Pharo, showing that the approach is general enough to be applied to other languages.

International audience; This paper proposes a new method to solve the monitoring and anomaly detection problems of Low-power Internet of Things (IoT) devices. However, their performances are constrained by limited processing, memory, and communication, usually using battery-powered energy. Polling driven mechanisms for monitoring the security, performance, and quality of service of these networks should be efficient and with low overhead, which makes it particularly challenging. The present work proposes the design of a novel method based on a Deep Reinforcement Learning (DRL) algorithm coupled with an Unsupervised Learning reward technique to build a pooling monitoring of IoT networks. This combination makes the network more secure and optimizes predictions of the DRL agent in adaptive environments.

International audience; Nowadays, many research studies and industrial investigations have allowed the integration of the Internet of Things (IoT) in current and future networking applications by deploying a diversity of wireless-enabled devices ranging from smartphones, wearables, to sensors, drones, and connected vehicles. The growing number of IoT devices, the increasing complexity of IoT systems, and the large volume of generated data have made the monitoring and management of these networks extremely difficult. Numerous research papers have applied Reinforcement Learning (RL) and Deep Reinforcement Learning (DRL) techniques to overcome these difficulties by building IoT systems with effective and dynamic decision-making mechanisms, dealing with incomplete information related to their environments. The paper first reviews pre-existing surveys covering the application of RL and DRL techniques in IoT communication technologies and networking. The paper then analyzes the research papers that apply these techniques in wireless IoT to resolve issues related to routing, scheduling, resource allocation, dynamic spectrum access, energy, mobility, and caching. Finally, a discussion of the proposed approaches and their limits is followed by the identification of open issues to establish grounds for future research directions proposal.

CNRS 2, FNEGE 2, HCERES A; International audience; Available technologies in organizations can be exploited for creating useful innovative solutions in a rather quick, efficient and economical way. Despite the huge benefits that can be derived by proactively exploiting available technologies and systems for innovation, there is relatively less academic research devoted to the subject. Motivated by this perceptible gap in the digital innovation literature, we theorize and examine the role of one possible structured creative mechanism (design thinking approach) for exploiting existing digital technologies to facilitate innovation. Grounding our arguments in the 'effective use of IS' and 'technology affordance' literature, we conceptualize innovation agent attributes and actualized innovation affordance achieved through the use of ICT, as the two salient facilitators for digital innovation. Innovation agent attributes are operationalized via domain generic and domain specific attributes of 'creative self-efficacy' and 'affect towards ICT use' and actualized innovation affordance is operationalized as the effective use of available technologies through an 'ICT enabled design-thinking approach'. The hypothesized model is empirically tested via a quasi-experimental method comprising data from a two-wave temporally separated survey. Results establish the significant influence of both-innovation agent attributes and actualized innovation affordance for facilitating digital innovation. Moreover, actualized innovation affordance, operationalized through ICT-DT, has a stronger association with digital innovation compared to the innovation agent attributes. This result underlines the need for organizations to proactively focus on using ICT effectively through structured creativity approaches (such as DT) for facilitating innovative outcomes. The study also offers several implications for research and practice.; Les technologies de l’information (TIC) disponibles dans les organisations peuvent être utilisées pour créer des solutions innovantes d’une manière rapide, efficace et économique. Malgré la nécessité de mieux comprendre les modalités permettant d’utiliser les TIC dans les organisations pour favoriser l’innovation, la littérature sur l’innovation numérique/digitale se concentre principalement sur l’étude de la mise en œuvre de TIC existantes dans de nouveaux contextes ou sur la conception et le développement de nouvelles technologies de l’information. Cet article enrichit ces travaux de recherche en étudiant un mécanisme créatif alternatif mobilisant les technologies digitales existantes pour l’innovation. En ancrant nos arguments dans la littérature sur l’usage efficace des SI et « l’affordance de la technologie », nous conceptualisons les caractéristiques des agents d’innovation et l’affordance actualisée de l’innovation basée sur l’utilisation des TIC, comme les deux principaux facilitateurs de l’innovation digitale. Les caractéristiques des agents d’innovation sont opérationnalisées par le biais de caractéristiques génériques et spécifiques au domaine de « l’auto-efficacité créative » et de « l’affect envers l’utilisation des TIC ». L’affordance de l’innovation actualisée est opérationnalisée comme l’utilisation efficace des technologies disponibles grâce à une approche de conception basée sur les TIC ou encore « ICT enabled design-thinking » (ICT-DT). Le modèle proposé est testé empiriquement via une méthode quasi-expérimentale comprenant des données provenant d’une enquête en deux temps de collecte. Les résultats montrent une influence significative à la fois des caractéristiques de l’agent d’innovation et de l’affordance actualisée de l’innovation pour faciliter l’innovation digitale. Cependant, l’affordance actualisée de l’innovation, opérationnalisée par l’approche ICT-DT, a une relation plus forte avec l’innovation digitale par rapport aux caractéristiques de l’agent d’innovation. Cela souligne la nécessité pour les organisations d’intégrer l’utilisation des TIC pour permettre des approches de créativité structurées (telles que le DT) et encourager les productions innovantes. L’étude offre également plusieurs implications pour la recherche et pour la pratique.

The Angular Resolution Limit (ARL) is a fundamental statistical metric to quantify our ability to resolve two closely-spaced narrowband far-field complex sources. This statistical quantity, is defined as the minimal angular deviation between the two sources to be separated for a prefixed detection-based performance. In this work, we assume that the sources of interest are corrupted by a compound-Gaussian noise. In the standard literature, denoting with δ the true distance between the two sources, the derivation of the ARL is based on the statistical distribution of the Generalized Likelihood Ratio Test (GLRT) for a binary test where there is only one source under the null hypothesis (i.e., δ = 0 ) and two sources under the alternative hypothesis δ ≠ 0. In literature, the true angular distance (TAD) is generally considered as an unknown deterministic parameter, then a maximum likelihood-based estimation of δ is exploited in the GLRT. In this paper, breaking away from existing contributions, we suppose that the TAD is a random variable, Gaussian distributed, meaning that δ ∼ N ( δ 0 , σ δ 2 ) . The TAD uncertainty can have many causes as for instance moving sources or/and platform, antenna calibration error, etc. In this work, a generic and flexible (but common) statistical model of the uncertain knowledge of the TAD is preferred instead of a too much specified error model. The degree of randomness (or uncertainty) is quantified by the ratio ξ = δ 0 2 / σ δ 2 . The standard framework of the GLRT is no longer feasible for our problem formulation. To cope with the compound Gaussian noise modeling and the random model of the TAD, a powerful upper bound from information/geometry theory is exploited in this work. More precisely, a new expected Chernoff Upper Bound (CUB) on the minimal error probability is introduced. Based on the analysis of this upper bound, we show that the expected-CUB is highly dependent on the degree of uncertainty, ξ. As a by-product, the optimal s-value in the Chernoff divergence for which the expected-CUB is the tightest upper bound is analytically studied and the role of the mean value δ0 in the ARL context is analyzed.

International audience; This paper proposes a novel approach based on adaptive dimensionality reduction (ADR) and a semi-supervised 3-D convolutional neural network (3-D CNN) for the spectro-spatial classification of hyperspectral images (HSIs). It tackles the problem of curse of dimensionality and the limited number of training samples by selecting the most relevant spectral bands. The selected bands should be informative, discriminative and distinctive. They are fed into a semi-supervised 3-D CNN feature extractor, then a linear regression classifier to produce the classification map. In fact, the proposed semi-supervised 3-D CNN model seeks to extract the deep spectral and spatial features based on convolutional encoder-decoder to enhance the HSI classification. It uses several 3-D convolution and max-pooling layers to extract these features from the selected relevant bands. The main advantage of the proposed approach is to reduce the high dimensionality of HSI, preserve the relevant spectro-spatial information and enhance the classification using few labeled training samples. Experimental studies are carried out on three real HSI data sets: Indian Pines, Pavia University, and Salinas. The obtained results show that the proposed approach performs better than other deep learning-based methods including CNN-based methods, and significantly improves the classification accuracy of HSIs.

Internet of Mobile Things (IoMT) is a new paradigm of the Internet of Things (IoT) where devices are inherently mobile. While mobility enables innovative applications and allows new services, it remains a challenging issue as it causes disconnection of nodes and intermittent connectivity, which negatively impacts the network operation and performance. In addition, energy saving presents a real challenge for networks with limited resources. In this context, a new energy efficient and mobility aware routing protocol called EMA-RPL is proposed based on the well-known Routing Protocol for Low power and Lossy Networks (the RPL standard). Unlike the RPL which is basically designed for static devices, EMA-RPL enables better sustaining of connectivity of mobile nodes and conserving energy. The proposed protocol integrates an enhanced mobility detection method through a continuous control of the distance between the mobile node and its attachment, a novel point of attachment prediction based on the new location of the mobile node, and an efficient replacement strategy preserving the mobile node energy. EMA-RPL overcomes and mitigates problems caused by the mobility of nodes. Simulations using Cooja/Contiki show that EMA-RPL outperforms both the RPL and its mobility aware variant (MRPL) in terms of handover delay, data loss rate, signaling cost and energy consumption.

International audience; Internet of Things (IoT) is increasingly deployed in different domains and environments including smart homes, smart cities, healthcare, industry 4.0, and smart agriculture, by connecting a large number of physical objects to deliver a new class of applications. The rising number of these connected objects and their heterogeneity have raised new research directions and challenges regarding their communications, scalability and the large amount of data that generate. As a result, new communication technologies have been proposed to be applied in these environments and applications, mainly to consider their main inherent properties which are the information that generate and handle, and the content that disseminate. Among the adopted techniques and recently integrated into the IoT is the ICN (Information Centric Networking) paradigm. The choice of integration of ICN in the context of IoT is mainly motivated by all the advantages it represents, in particular content caching and decoupling senders and receivers. In this paper, we provide a detailed analysis of the motivations behind using ICN in IoT environments and a survey of existing research work that have already applied ICN as a communication support for IoT applications.

International audience; Big data analysis assumes a significant role in Earth observation using remote sensing images, since the explosion of data images from multiple sensors is used in several fields. The traditional data analysis techniques have different limitations on storing and processing massive volumes of data. Besides, big remote sensing data analytics demand sophisticated algorithms based on specific techniques to store to process the data in real-time or in near real-time with high accuracy, efficiency, and high speed. In this paper, we present a method for storing a huge number of heterogeneous satellite images based on Hadoop distributed file system (HDFS) and Apache Spark. We also present how deep learning algorithms such as VGGNet and UNet can be beneficial to big remote sensing data processing for feature extraction and classification. The obtained results prove that our approach outperforms other methods.

International audience

Drought disasters significantly affected human life and water resources. Therefore, forecasting methods like statistical models, machine learning, and deep learning architectures help scientists to take effective decisions to decrease the effects of natural disasters by providing decision-making plans. Droughts can be forecasted using meteoro-logical indices like the standardized precipitation evapotranspiration index (SPEI), which aid governments in taking drought-prevention steps. In this paper, we present a big drought architecture for drought modeling and forecasting. The proposed architecture is composed of 5 layers: Data collection, data preprocessing, data storage, data processing and interpretation, and decision making. Besides, we present a comparative study between three different methods ARIMA, PROPHET, and LSTM for drought forecasting. Three different metrics are used for the performance evaluation Root Mean Squared Error (RMSE), coefficient of determination (R2), and Mean Squared Error (MAE). Experiments are carried out using data from the province of Jiangsu. Results revealed that LSTM outperformed the other models, and ARIMA outperformed the PROPHET model.

International audience; This paper deals with the problem of prescribedtime stabilization of controllable linear systems with input delay. The problem is reformulated under a cascade PDE-ODE setting from which a prescribed-time predictor feedback is designed based on the backstepping approach, and whose transformation makes use of time-varying kernels. The bounded invertibility of the transformation is guaranteed. It is proved that the solution converges to the equilibrium in a prescribed-time. A simulation example is presented to illustrate the results.

International audience; The nowadays CPPS (Cyber Physical Production System) depends essentially on Operational Technologies (OTs) that have been acquired from a wide range of Information Technologies (ITs). In order to guarantee the CPPSs requirements, several of ITs have been adjusted or adapted. As a result, new issues appeared, in the design of this production system, such as the integration of safety, security and vulnerability risks. The greater part of CPPSs vulnerabilities have generally not been considered in the design phase. However, this dependence on ITs makes CPPSs increasingly vulnerable to cyber-attacks and security threats, which affect their global performance. The focus of this research is to address the cyber security challenges that these systems have to cope. It discusses the features of the CPPSs threat environment and align safety and security risk analysis. In fact, the first problem in security risk analyzes is determining the likelihood of the cyber-attacks. The use likelihood in security analysis is not practical and sometimes does not make sense. Therefore, this prevents the research community from using a panoply of methods that exist in the field of safety because they are based on the use of probabilities. In this paper it is introduced a new vulnerability scoring system for industrial systems, called ICVSS. The proposed approach based on the calculation of variety of metrics distributed into two groups of metrics inherited from the classical CVSS metrics. The ICVSS not only make possible to assess the vulnerability and to ease the communication between the safety and security teams, but also it is a good alternative to replace the likelihood in order to be able to use these safety methods. Next, a new approach to co-analysis and co-assessment of safety and security called FVMEARA (Failure and Vulnerability Modes, and Effect Analysis and Risk Assessment) is presented. Actually, FVMEARA is a use case that shows the effectiveness of our methodology in which we have been able to reuse the FMEA method, known as the most widely used in the field of safety, to carry out a co-analysis of hazards and threats and to assess their risks. Even more, the proposed solution complies with the standards IEC 60812 for safety and the standard IEC 62443 for security.

International audience; Low-power Internet of Things (IoT) networks are widely deployed in various environments with resource constrained devices, making their states monitoring particularly challenging. In this paper, we propose an adaptive monitoring mechanism for low-power IoT devices, by using a reinforcement learning (RL) method to automatically adapt the polling frequencies of the collected attributes. Our goal is to minimize the number of monitoring packets while keeping accurate and timely detection of threshold crossings associated to supervised attributes. We study the various RL parameter settings under different monitoring attribute behaviors using OpenAi Gym simulator. We implement the RL based adaptive polling in Contiki OS and we evaluate its performance using Cooja simulator. Our results show that our approach converges to optimal polling frequencies and outperforms static periodic notification-based methods by reducing the number of monitoring packets, with a percentage of correctly detected threshold crossings exceeding 80%.

A robust nonlinear control is designed for stabilizing linear multi-input–multi-output systems. The presented control law homogenizes a linear system (without its transformation to a canonical form) with a specified degree and stabilizes it in a finite time (or with a fixed-time attraction to any compact set containing the origin) if the degree of homogeneity is negative (positive). The tuning procedure is formalized in an linear matrix inequalities (LMI) form. Performance of the approach is illustrated by numerical and experimental examples.

International audience; High accuracy air pollution monitoring in a smart city requires the deployment of a huge number of sensors in this city. One of the most appropriate wireless technologies expected to support high density deployment is LoRaWAN which belongs to the Low Power Wide Area Network (LPWAN) family and offers long communication range, multi-year battery lifetime and low cost end devices. It has been designed for End Devices (EDs) and applications that need to send small amounts of data a few times per hour. However, a high number of end devices breaks the orthogonality of LoRaWAN transmissions, which was one of the main advantages of LoRaWAN. Hence, network performances are strongly impacted. To solve this problem, we propose a solution called OAPM (Orthogonal Air Pollution Monitoring) which ensures the orthogonality of LoRaWAN transmissions and provides accurate air pollution monitoring. In this paper, we show how to organize EDs into clusters and sub-clusters, assign transmission times to EDs, configurate and synchronize them, taking into account the specificities of LoRaWAN and the features of the air pollution monitoring application. Simulation results corroborate the very good behavior of OAPM.

The ultimate goal of this project is to develop a recommender system for the SmartVideo platform. The platform streams different content of local channels for the Grand Est Region of France to a large public. So, we aim to propose a solution to alleviate the data representation and data collection issue of recommender systems by adopting and adjusting the xAPI standard to fit our case of study and to be able to represent our usage data in a formal and consistent format. Then, we will propose and implement a bunch of recommendation algorithms that we are going to test in order to evaluate our developed recommender system., Le but ultime de ce projet est de développer un système de recommandation dédié à la plateforme SmartVideo de diffusion de vidéo en ligne. En effet, la plateforme met à disposition diverses contenus des chaînes locales de la région Grand Est du France. Alors, nous allons présenter une solution pour alléger le problème de représentation et de collecte de données d’usages par adopter et ajuster le standard xAPI pour représenter et collecter les données de façon simple et formelle. Ensuite, nous allons proposer et implanter des algorithmes de recommandation que nous allons les tester pour évaluer notre système de recommandation.

With the rapid growth of image collections, image classification and annotation has been active areas of research with notable recent progress. Bag-of-Visual-Words (BoVW) model, which relies on building visual vocabulary, has been widely used in this area. Recently, attention has been shifted to the use of advanced architectures which are characterized by multi-level processing. Hierarchical Max-Pooling (HMAX) model has attracted a great deal of attention in image classification. To improve image classification and annotation, several approaches based on ontologies have been proposed. However, image classification and annotation remain a challenging problem due to many related issues like the problem of ambiguity between classes. This problem can affect the quality of both classification and annotation results. In this paper, we propose an ontology-based image classification and annotation approach. Our contributions consist of the following: (1) exploiting ontological relationships between classes during both image classification and annotation processes; (2) combining the outputs of hypernym–hyponym classifiers to lead to a better discrimination between classes; and (3) annotating images by combining hypernym and hyponym classification results in order to improve image annotation and to reduce the ambiguous and inconsistent annotations. The aim is to improve image classification and annotation by using ontologies. Several strategies have been experimented, and the obtained results have shown that our proposal improves image classification and annotation.

International audience; The Internet of Things (IoT) ecosystem includes a plethora of devices equipped with heterogeneous communication interfaces. They exhibit different mobility patterns and hardware constraints as memory, battery and processing power. On the other hand, IoT applications are overlayed on top of these constrained-devices imposing stringent requirements in term of data availability, data coherence, and response latency. To cope with these challenges, Named Data Networking (NDN) architecture is positioned in the middle layer to act as the networking layer. By providing easy data access thanks to the unique and location-independent content names, in-network caching and name-based routing, NDN expects to hide from IoT applications the complexity and diversity of the underlying Things by adapting the network operation to their features. In this paper, we focus on data availability requirements threatened by high IoT network dynamics related to sensors mobility. We address the producer mobility issue in NDN/IoT networks using the routing-based approach. We propose a novel and efficient forwarding algorithm named AFIRM in order to support producer mobility, and compare its performances to those of other relevant solutions based on the routing approach.

BackgroundAn important but rarely addressed question in nano-therapy is to know whether bio-degraded nanoparticles with reduced sizes and weakened heating power are able to maintain sufficient anti-tumor activity to fully eradicate a tumor, hence preventing tumor re-growth. To answer it, we studied magnetosomes, which are nanoparticles synthesized by magnetotactic bacteria with sufficiently large sizes (~ 30 nm on average) to enable a follow-up of nanoparticle sizes/heating power variations under two different altering conditions that do not prevent anti-tumor activity, i.e. in vitro cellular internalization and in vivo intra-tumor stay for more than 30 days.ResultsWhen magnetosomes are internalized in U87-Luc cells by being incubated with these cells during 24 h in vitro, the dominant magnetosome sizes within the magnetosome size distribution (DMS) and specific absorption rate (SAR) strongly decrease from DMS ~ 40 nm and SAR ~ 1234 W/gFebefore internalization to DMS ~ 11 nm and SAR ~ 57 W/gFe after internalization, a behavior that does not prevent internalized magnetosomes to efficiently destroy U87-Luc cell, i.e. the percentage of U87-Luc living cells incubated with magnetosomes decreases by 25% between before and after alternating magnetic field (AMF) application. When 2 µl of a suspension containing 40 µg of magnetosomes are administered to intracranial U87-Luc tumors of 2 mm3and exposed (or not) to 15 magnetic sessions (MS), each one consisting in 30 min application of an AMF of 27 mT and 198 kHz, DMS and SAR decrease between before and after the 15 MS from ~ 40 nm and ~ 4 W/gFedown to ~ 29 nm and ~ 0 W/gFe. Although the magnetosome heating power is weakened in vivo, i.e. no measurable tumor temperature increase is observed after the sixth MS, anti-tumor activity remains persistent up to the 15th MS, resulting in full tumor disappearance among 50% of treated mice.ConclusionHere, we report sustained magnetosome anti-tumor activity under conditions of significant magnetosome size reduction and complete loss of magnetosome heating power.

International audience; In the context of software product line engineering, collaborative configuration is a decision-making process where multiple stakeholders contribute in building a single product specification. Several approaches addressing collaboration during configuration have already been proposed, but we still have little hard evidence about their effectiveness and little understanding about how collaborative configuration process should be carried out. This paper presents a classification framework to help understand existing collaborative configuration approaches. To elaborate it, a systematic mapping study was conducted guided by three research questions and 41 primary studies was selected out of 238 identified ones. The proposed framework is composed of four dimensions capturing main aspects related to configuration approaches: purpose, collaboration, process and tool. Each dimension is itself multi-faceted and a set of attributes is associated to each facet. Using this framework, we position and classify existing approaches, structure the representation of each approach characteristics, highlight their strengths and weaknesses, compare them to each other, and identify open issues. This study gives a solid foundation for classifying existing and fu-* Corresponding author ture approaches for product lines collaborative configuration. Researchers and practitioners can use our framework for identifying existing research/technical gaps to attack, better scoping their own contributions, or understanding existing ones.

International audience; In the context of software product line engineering, collaborative configuration is a decision-making process where multiple stakeholders contribute in building a single product specification. Several approaches addressing collaboration during configuration have already been proposed, but we still have little hard evidence about their effectiveness and little understanding about how collaborative configuration process should be carried out. This paper presents a classification framework to help understand existing collaborative configuration approaches. To elaborate it, a systematic mapping study was conducted guided by three research questions and 41 primary studies was selected out of 238 identified ones. The proposed framework is composed of four dimensions capturing main aspects related to configuration approaches: purpose, collaboration, process and tool. Each dimension is itself multi-faceted and a set of attributes is associated to each facet. Using this framework, we position and classify existing approaches, structure the representation of each approach characteristics, highlight their strengths and weaknesses, compare them to each other, and identify open issues. This study gives a solid foundation for classifying existing and fu-* Corresponding author ture approaches for product lines collaborative configuration. Researchers and practitioners can use our framework for identifying existing research/technical gaps to attack, better scoping their own contributions, or understanding existing ones.

International audience; The Information-Centric Networking (ICN) paradigm is shaping the foreseen future Internet architecture by focusing on the data itself rather than its hosting location. It is a shift from a host-centric communication model to a content-centric model supporting among others unique and location-independent content names, in-network caching and name-based routing. By leveraging the easy data access, and reducing both the retrieval delay and the load on the data producer, the ICN can be a viable framework to support the Internet of Things (IoT), interconnecting billions of heterogeneous constrained objects. Among several ICN architectures, the Named Data Networking (NDN) is considered as a suitable ICN architecture for IoT systems. However, its default caching approach lacks a data freshness mechanism, while IoT data are transient and frequently updated by the producer which imposes stringent requirements in terms of information freshness. Furthermore, IoT devices are usually resource-constrained with harsh limitations on energy, memory and processing power. We propose in this paper a caching strategy and a novel cache freshness mechanism to monitor the validity of cached contents in an IoT environment while minimizing the caching process cost. We compared our solution to several relevant schemes using the ccnSim simulator. Our solution exhibits the best system performances in terms of hop reduction ratio, server hit reduction ratio and response latency, yet it provides the lowest cache cost and significantly improves the content validity.

International audience; An objective of satellite remote sensing is to predict or characterize the land cover change (LCC) over time. Sometimes we are capable of describing the changes of land cover with a probability distribution. However, we need sufficient knowledge about the natural variability of these changes, which is not always possible. In general, uncertainties can be subdivided into aleatory and epistemic. The main problem is that classical probability theory does not make a clear distinction between aleatory and epistemic uncertainties in the way they are represented, i.e., both of them are described with a probability distribution. The aim of this paper is to propagate the aleatory and epistemic uncertainty associated with both input parameters (features extracted from satellite image object) and model structure of LCC prediction process using belief function theory. This will help reducing in a significant way the uncertainty about future changes of land cover. In this study, the changes prediction of land cover in Cairo region, Egypt for next 16 years (2030) is anticipated using multi-temporal Landsat TM5 satellite images in 1987 and 2014. The LCC prediction model results indicated that 15% of the agriculture and 6.5% of the desert will be urbanized in 2030. We conclude that our method based on belief function theory has a potential to reduce uncertainty and improve the prediction accuracy and is applicable in LCC analysis.

Image hash is an alternative to cryptographic hash functions for checking integrity of digital images. Compared to cryptographic hash functions, an image hash or a Perceptual Hash Function (PHF) is resilient to content preserving distortions and sensitive to malicious tampering. In this paper, a robust and secure image hashing technique using a Gaussian pyramid is proposed. A Gaussian pyramid decomposes an image into different resolution levels which can be utilized to obtain robust and compact hash features. These stable features have been utilized in the proposed work to construct a secure and robust image hash. The proposed scheme uses Laplacian of Gaussian (LOG) and disk filters to filter the low-resolution Gaussian decomposed image. The filtered images are then subtracted and their difference is used as a hash. To make the hash secure, a key is introduced before feature extraction, thus making the entire feature space random. The proposed hashing scheme has been evaluated through a number of experiments involving cases of non-malicious distortions and malicious tampering. Experimental results reveal that the proposed hashing scheme is robust against non-malicious distortions and is sensitive to detect minute malicious tampering. Moreover, False Positive Probability (FPP) and False Negative Probability (FNP) results demonstrate the effectiveness of the proposed scheme when compared to state-of-the-art image hashing algorithms proposed in the literature.

International audience; Product line engineering is a new production paradigm that provides organizations a competitive edge by improving productivity and decreasing costs. The purpose with this new production paradigm is no longer to develop a single product but to develop a product family and to generate the products of the line through configuration processes. However, the potential benefits of product line engineering can be missed when dealing with large product lines because the configuration processes become error prone tasks. Consequently, guiding stakeholders during such complex configuration processes and recommending the best configuration alternatives until leading to a satisfying experience becomes a challenge. This paper focuses on enhancing product line configuration processes through process mining techniques. Therefore, user's actions of previous product line configurations are logged, mined and analyzed. We conducted a preliminary research to motivate the advantages of process mining in product line configuration and to explore what can process mining bring for configuration processes and how to use it to enhance configuration processes. Thus, guidance questions are sketched in order to position process mining as a solving tool for the configuration difficulties. Furthermore, we propose a reference architecture that considers process mining for configuring product lines.

A new hyperspectral texture descriptor, Relative Spectral Difference Occurrence Matrix (RSDOM) is proposed. Developed in a metrological framework, it simultaneously considers the distribution of spectra and their spatial arrangement in the hyperspectral image. It is generic and adapted for any number of spectral band or range. As validation, a texture classification scheme is applied on HyTexiLa dataset using RSDOM. The obtained accuracy is excellent (95.6%), comparable to Opponent Band Local Binary Pattern (OBLBP) but at a much-reduced feature size (0.1% of OBLBP's). © 2019 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

International audience; Robust calibration of next-generation radio-interferometers, as the square kilometer array (SKA) for instance, is a crucial prepro-cessing step for sky imaging. Recently, several robust calibration es-timators based on the use of well known strong sources in the field of view (FOV) have been proposed in the literature. For that, usually a compound-Gaussian (CG) noise is considered since it can take into account the presence of outliers, such that unknown weak sources in the FOV, and remains mathematically tractable. Specifically, the CG model is a zero-mean Gaussian process with a random variance, usually called texture. Performance bounds provide the lowest mean squared error (MSE) that an unbiased estimator can hope to reach. To the best of our knowledge, there is a lack of practical bounds dedicated to the robust calibration of future radio interferometers as far as they can be used as design tools for a given FOV. In this work, the hybrid Cramér-Rao bound (HCRB) is derived for a propagation model based on the Jones-matrix formalism corrupted by a CG noise for several texture distributions (K-distribution, Student's t, Cauchy, and Inverse-Gaussian compound Gaussian distribution (IG-CG)). In this standard physical model, the waveform propagation model is parametrized by a set of deterministic physical parameters of interest (complex gains, phases, DOA, etc.). In this work, we show that the HCRB w.r.t. the physical deterministic parameters is in fact given by the modified CRB (MCRB). The MCRB is easy to derive since only the first-order moment of the inverse of the texture variable is needed.

With the increase in measurement/sensing technologies, the collected data are in intrinsically multidimensional in a large number of applications. This can be interpreted as a growth of the dimensionality/order of the associated tensor. There exists therefore a crucial need to derive equivalent and alternative models of a high-order tensor as a graph of low-order tensors. In this work we consider a " train " graph, i.e., a Q-order tensor will be represented as a Tensor Train (TT) composed of Q − 2 3-order core tensors and two core matrices. In this context, it has been shown that a canonical rank-R CPD / PARAFAC model can always be represented exactly by a TT model whose cores are canonical rank-R CPD/PARAFAC. This model is called TT-CPD. We generalize this equivalence to the PARALIND model in order to take into account potential linear dependencies in factors. We derive and discuss here uniqueness conditions for the case of the TT-PARALIND model., Avec l'accroissement des moyens d'acquisition et de mesures, les données d'intérêt sont par essence de nature multidimensionnelle. Ceci peut s'interpréter comme un accroissement de la dimension/ordre des modèles tensoriels associés. Il y a donc un besoin crucial d'avoir à notre disposition des représentations équivalentes d'un tenseur d'ordre élevé en un graphe de tenseurs d'ordre réduit. Dans ce travail nous considérons un graphe de type "train", c'est-à-dire, qu'un tenseur d'ordre Q sera représenté par un train de tenseur (TT) composé de Q − 2 coeurs tensoriels d'ordre 3 et deux coeurs matriciels. Dans ce cadre, il a été démontré qu'un modèle CPD/PARAFAC de rang canonique R peut être toujours représenté de manière exacte par un modèle TT dont les coeurs sont eux-même CPD/PARAFAC de rang canonique R. Ce modèle est nommé TT-CPD. Nous généralisons cette équivalence au modèle PARALIND afin de prendre en compte des potentielles dépendances linéaires dans les facteurs. Nous donnons et discutons ici les conditions d'unicité dans le cas du modèle TT-PARALIND.

International audience; Spaceborne and airborne sensors deliver a huge number of Earth Observation Data every day. In this context, we can easily observe the whole earth from its different sides. Therefore, this big data is important in remote sensing and could be exploited in several domains requiring image classification, natural hazard monitoring, global climate change, agriculture, urban planning. Over the last five years, Convolutional Neural Networks (CNN) emerged as the most successful technique for the image classification task, as well as a number of other computer vision tasks. However, to train millions of parameters in CNN one requires a huge amount of annotated data. This requirement leads to a significant challenge if the available training data is limited for a target task at hand. To address this challenge, in the recent literature, researchers proposed various ways to apply a technique called Transfer Learning to transfer the knowledge gained by training CNNs parameters on some large annotated dataset to the target task with limited availability of training data. Most of our work in this paper was dedicated to proposing a hybrid classification of remote sensing images. This architecture combines Spark RDD image coding to consider image's local regions, pre-trained VGGNET-16 and UNET for image segmentation and SVM (Support Vector Machines) from spark Machine Learning to achieve labeling task.

The explosion of web 2.0 and social networks has created an enormous and rewarding source of information that has motivated researchers in different fields to exploit it. Our work revolves around the issue of access and identification of social information and their use in building a user profile enriched with a social dimension, and operating in a process of personalization and recommendation. We study several approaches of Social IR (Information Retrieval), distinguished by the type of incorporated social information. We also study various social recommendation approaches classified by the type of recommendation. We then present a study of techniques for modeling the social user profile dimension, followed by a critical discussion. Thus, we propose our social recommendation approach integrating an advanced social user profile model., L’explosion du web 2.0 et des réseaux sociaux a crée une source d’information énorme et enrichissante qui a motivé les chercheurs dans différents domaines à l’exploiter. Notre travail s’articule autour de la problématique d’accès et d’identification des informations sociales et leur exploitation dans la construction d’un profil utilisateur enrichi d’une dimension sociale, et son exploitation dans un processus de personnalisation et de recommandation. Nous étudions différentes approches sociales de RI (Recherche d’Information), distinguées par le type d’informations sociales incorporées. Nous étudions également diverses approches de recommandation sociale classées par le type de recommandation. Nous exposons ensuite une étude des techniques de modélisation de la dimension sociale du profil utilisateur, suivie par une discussion critique. Ainsi, nous présentons notre approche de recommandation sociale proposée intégrant un modèle avancé de profil utilisateur social.

The explosion of web 2.0 and social networks has created an enormous and rewarding source of information that has motivated researchers in different fields to exploit it. Our work revolves around the issue of access and identification of social information and their use in building a user profile enriched with a social dimension, and operating in a process of personalization and recommendation. We study several approaches of Social IR (Information Retrieval), distinguished by the type of incorporated social information. We also study various social recommendation approaches classified by the type of recommendation. We then present a study of techniques for modeling the social user profile dimension, followed by a critical discussion. Thus, we propose our social recommendation approach integrating an advanced social user profile model. L’explosion du web 2.0 et des réseaux sociaux a crée une source d’information énorme et enrichissante qui a motivé les chercheurs dans différents domaines à l’exploiter. Notre travail s’articule autour de la problématique d’accès et d’identification des informations sociales et leur exploitation dans la construction d’un profil utilisateur enrichi d’une dimension sociale, et son exploitation dans un processus de personnalisation et de recommandation. Nous étudions différentes approches sociales de RI (Recherche d’Information), distinguées par le type d’informations sociales incorporées. Nous étudions également diverses approches de recommandation sociale classées par le type de recommandation. Nous exposons ensuite une étude des techniques de modélisation de la dimension sociale du profil utilisateur, suivie par une discussion critique. Ainsi, nous présentons notre approche de recommandation sociale proposée intégrant un modèle avancé de profil utilisateur social.

International audience; The increasing volume of data generated by earth observation programs such as Copernicus, NOAA, and NASA Earth Data, is overwhelming. Although these programs are very costly, data usage remains limited due to lack of interoperability and data linking. In fact, multi-source and heterogeneous data exploitation could be significantly improved in different domains especially in the natural disaster prediction one. To deal with this issue, we introduce the PREDICAT project that aims at providing a semantic service-oriented platform to PREDIct natural CATastrophes. The PREDICAT platform considers (1) data access based on web service technology; (2) ontology-based interoperability for the environmental monitoring domain; (3) data integration and linking via big data techniques; (4) a prediction approach based on semantic machine learning mechanisms. The focus in this paper is to provide an overview of the PREDICAT platform architecture. A scenario explaining the operation of the platform is presented based on data provided by our collaborators, including the international intergovernmental Sahara and Sahel Observatory (OSS).

International audience; Multi-source observed data are generally characterized by their syntactic, structural and semantic heterogeneities. A key challenge is the semantic interoperability of these data. In this context, we propose an ontology-based system that supports environmental monitoring. Our contributions could be resumed around 1) the construction of an ontology which allows to represent the knowledge and reuse it in a real-world way, 2) the guarantee of the semantic interoperability of ontological modules since the proposed ontology is based on the upper level ontology Basic Formal Ontology (BFO) 3) the modularity of the proposed ontology in order to facilitate its reuse and evolution. The proposed ontology has been implemented and evaluated using quality metrics. We also present a real use case study that demonstrates how the proposed ontology allows implicit knowledge generation.

International audience; In this paper, we design and experiment ODYSSE (Opportunistic Duty cYcle based routing protocol for wirelesS Sensor nEtworks) protocol. It combines three main mechanisms: (i) duty cycle, where nodes alternate between active and sleep states, (ii) opportunistic routing where routing tables do not exist and the next hop is elected once the packet arrives, and (iii) source coding with LDPC (Low-Density Parity-Check) codes in order to face packet losses while minimizing energy consumption. We focus on two heterogeneous scenarios: bulk image transmission and infrequent events reporting. Modeling the average waiting delay of forwarders, we show that simple relay selection strategies are effective. We used 45 Arduino nodes communicating with IEEE 802.15.4 (XBee) within the large platform FIT IoT-LAB (IoT-LAB is part of the large platform FIT: Future Internet of Things). We implement and extensively study the behavior and performance of our proposal ODYSSE. We show that the three techniques fit perfectly, yielding a robust low complexity protocol for highly constrained nodes in typical IoT applications.

International audience; This paper addresses the problem of human action recognition from sequences of 3D skeleton data. For this purpose, we combine a deep learning network with geometric features extracted from data lie on a non-Euclidean space, which have been recently shown to be very effective to capture the geometric structure of the human pose. In particular, our approach claims to incorporate the intrinsic nature of the data characterized by Lie Group into deep neural networks and to learn more adequate geometric features for 3D action recognition problem. First, geometric features are extracted from 3D joints of skeleton sequences using the Lie group representation. Then, the network model is built from stacked units of 1-dimensional CNN across the temporal domain. Finally, CNN-features are then used to train an LSTM layer to model dependencies in the temporal domain, and to perform the action recognition. The experimental evaluation is performed on three public datasets containing various challenges: UT-Kinect, Florence 3D-Action and MSR-Action 3D. Results reveal that our approach achieves most of the state-of-the-art performance.

Data lakes have recently emerged as an alternative solution to costly traditional data warehouse solutions. To exploit data lakes, however, there is a need for means that assist users in combining and integrating data stored within a data lake. In this paper, we position ourselves in the recurrent context where a user has a local dataset that is not sufficient for processing the queries that are of interest to him/her. We show how data lakes, or more specifically the service lakes, since we are focusing on data providing services, can be leveraged to answer user queries, taking into account the quality of the services and respecting the (time and monetary) budget set by the user.

International audience; Leishmania (L.) infantum-infected dogs may present with a large range of clinical signs, from apparently healthy with no or few (asymptomatic dogs, AD) to several clinical signs indicators of active infection (symptomatic dogs, SD). The present study is justified by the conflicting reports describing that either L. infantum-specific IgG1 or IgG2 antibodies may be used as isotype marker of the asymptomatic infection status and by the lack, to our knowledge, of previous analysis of the IgG sub-classes autoantibody repertoires of Leishmania-infected dogs. On the basis of clinical evaluation and laboratory testing (IFAT, parasitological examination of Giemsa-stained lymph node smears, L. infantum antigens-ELISA of total (Tot) IgG), 131 dogs were categorized as SD, asymptomatic seronegative (AND) or seropositive dogs (APD) from surrounding areas, and as negative control dogs (CTD). ELISA based on leishmanial native antigens or recombinant LACK and LeIF proteins showed that SD produce higher levels of specific Tot IgG, IgG1 and IgG2 antibodies than APD, and that for both clinical stages, the antibody titers of IgG2 isotype were constantly higher than those of the IgG1. The seroprevalences of Tot IgG, IgG2 did not differ between APD and SD groups (97 and 97% in SD; 100 and 96% in APD, respectively) whereas that of IgG1 was slightly lower in SD (88% of APD versus 82% of SD). The autoantibody repertoires were analyzed by ELISA using HEp-2 extracts, ds-DNA, human albumin and transferrin as self-antigens and by Western blot using HEp-2 proteins. ELISA results' indicated that APD develop higher levels of IgG1 autoantibodies, and higher seroprevalence (50% and 26% in APD and SD, respectively), contrasting with lower levels and seroprevalences of Tot IgG and IgG2 (43 and 68% for APD; 100 and 74% for SD). Interestingly, SD showed a stronger IgG1 and particularly IgG2 reactivity with transferrin, an iron-binding protein, than APD and AND. Western blotting experiments produced heterogeneous IgG1 and IgG2 inter- and intra-groups reactivity profiles towards HEp-2 proteins, to identify a specific antigenic profile. Generated data from competitive HEp-2-ELISA using leishmanial antigens as inhibitors were in favor that IgG1 antibodies are predominantly autoantibodies to self-antigens in APD whereas they are mainly cross-reactive (Leishmania/self-antigens) in SD.

An online reinforcement learning algorithm is anytime if it does not need to know in advance the horizon T of the experiment. A well-known technique to obtain an anytime algorithm from any non-anytime algorithm is the "Doubling Trick". In the context of adversarial or stochastic multi-armed bandits, the performance of an algorithm is measured by its regret, and we study two families of sequences of growing horizons (geometric and exponential) to generalize previously known results that certain doubling tricks can be used to conserve certain regret bounds. In a broad setting, we prove that a geometric doubling trick can be used to conserve (minimax) bounds in $R_T = O(\sqrt{T})$ but cannot conserve (distribution-dependent) bounds in $R_T = O(\log T)$. We give insights as to why exponential doubling tricks may be better, as they conserve bounds in $R_T = O(\log T)$, and are close to conserving bounds in $R_T = O(\sqrt{T})$.; Un algorithme en ligne d'apprentissage par renforcement est dit "à tout moment" (anytime) s'il n'a pas besoin de connaître à l'avance l'horizon T de l'expérience. Une technique bien connue pour obtenir un algorithme à tout moment à partir d'un algorithme qui ne l'est pas est "l'astuce de doublement" (Doubling Trick). Dans le contexte des bandits multi-bras adverses ou stochastiques, la performance d'un algorithme est mesurée par son regret, et nous étudions deux familles de séquences d'horizons croissants (géométrique et exponentielle), pour généraliser des résultats précédemment connus que certaines astuces de doublement peuvent être utilisées pour conserver certaines limites de regret. Dans un cadre très générique, nous prouvons qu'une astuce géométrique de doublement peut être utilisée pour conserver les bornes (minimax) en $R_T = O(\sqrt{T})$ mais ne peut pas conserver les bornes (dépendantes de la distribution) en $R_T = O(\log T)$. Nous donnons un aperçu des raisons pour lesquelles les astuces de doublage exponentiel peuvent être meilleures, car elles conservent les bornes en $R_T = O(\log T)$, et sont proches de conserver les bornes en $R_T = O(\sqrt{T}$).

International audience; Hybrid clouds are increasingly used to outsource non-critical applications to public clouds. However, the main challenge within such environments, is to ensure a cost-efficient distribution of the systems between the resources that are on/off premises. For Multi Agent Systems (MAS), this challenge is deepened due to irregular workload progress and intensive communication between the agents, which may result in high computing and data transfer costs. Thus, in this paper we propose a generic framework for adaptive cost-efficient deployment of MAS with a special focus on hybrid clouds. The framework is based mainly on the use of a performance evaluation process that consists of simulating various partitioning options to estimate and optimize the overall deployment costs. Further, to cope with the irregular workload changes within a MAS and dynamically adapt its initial deployment, we propose an extended version of the Fiduccia-Mattheyses algorithm (E-FM). The experimental results highlight the efficiency of E-FM and show that an efficient MAS deployment to hybrid clouds depends on various factors such as the cloud providers and their different cost-models, the network state, the used partitioning algorithm, and the initial deployment.

International audience; A fundamental objective of remote sensing imagery is to spread out the knowledge about our environment and to facilitate the interpretation of different phenomena affecting the Earth's surface. The main goal of this chapter is to understand and interpret possible changes in order to define subsequently strategies and adequate decision-making for a better soil management and protection. Consequently, the semantic interpretation of remote sensing data, which consists of extracting useful information from image date for attaching semantics to the observed phenomenon, allows easy understanding and interpretation of such occurring changes. However, performing change interpretation task is not only based on the perceptual information derived from data but also based on additional knowledge sources such as a prior and contextual. This knowledge needs to be encoded in an appropriate way for being used as a guide in the interpretation process. On the other hand, interpretation may take place at several levels of complexity from the simple recognition of objects on the analyzed scene to the inference of site conditions and to change interpretation. This chapter presents semantic scenes analysis for change interpretation strategy in remote sensing imagery. The presented strategy is composed of different levels of interpretation. For each level, information elements such as data, information, and knowledge need to be represented and characterized. This chapter highlights the importance of ontologies exploiting for encoding the domain knowledge and for using it as a guide in the semantic scene interpretation task.

International audience; In this paper, we define a cluster based scheduling algorithm for Cognitive Radio Sensor Networks (CRSNs). To avoid inter-clusters collision, we assign fixed channels only to nodes having one-hop neighbors out of their clusters. We denote these nodes as specific nodes. Previous studies assign distinct channels to whole neighbor clusters to avoid inter-clusters collision. Our objective is to optimize the spatial reuse and to increase the network throughput while saving sensors energy. We start by assigning channels only to the specific nodes. Once the problem of inter-clusters collision is solved, each cluster head (CH) schedules the transmissions in its cluster independently. For the cluster members that are specific nodes, the CH assigns only time slots because the channel assignment is already done. For other cluster members (CMs) (not specific nodes), the CH assigns the pair (channel, slot). Two solutions are proposed in this paper to schedule the CMs: The Frame Intra Cluster Multichannel Scheduling algorithm denoted Frame-ICMS and the Slot Intra Cluster Multichannel Scheduling algorithm denoted Slot-ICMS. We evaluate the performance of these algorithms in case of accurate PUs activity detection and in case of bad PUs activity estimation. We prove that our proposals outperform an existing one especially in terms of energy saving.

Extended version of a paper presented at ACIT 2017. Full text available at https://iajit.org/portal/PDF/Special%20Issue%202018,%20No.%203A/17414.pdf; International audience; Syntactic and semantic resources play an important role for various Natural Language Processing (NLP) tasks by providing information about the correct structural representations of the sentences and their meaning. To date, there is not a wide-coverage electronic grammar for the Arabic language. In this context, we present a new approach for building a tree adjoining grammar to represent the syntax and the semantic of modern standard Arabic. This grammar is produced semi-automatically with the XMG (eXtensible MetaGrammar) description language. First the syntax of Arabic is described using the defined Arab-XMG meta-grammar. Then semantic information is added by introducing semantic frame-based dimension into the meta-grammar. This is achieved by exploiting lexical resources such as Arabic VerbNet. Finally, the link between semantic and syntax is established using a syntax-semantic interface that allows the construction of sentence meaning through semantic role labeling. Experiments were performed to check grammar coverage as well as the syntactic-semantic analysis. The results showed that the generated grammar can cover the basic syntactic structures of Arabic sentences and the different phrasal structures with a precision rate of about 92%. Moreover, it confirms the effectiveness of the proposed approach as we were able to parse semantically a set of sentences and build their semantic representations with a precision rate of about 72%.

International audience; Safety applications in Vehicular Ad-hoc Network (VANET) are used to reduce the number of accidents, which necessitates good Quality of Service (QoS), especially in some urgent cases and critical traffic. In fact, several Medium Access Control (MAC) protocols were proposed to provide high-performance for vehicular networks. For instance, the Contention-free MAC allows several vehicles to share the same communication channel. However, contention-based MAC protocols do not require any predefined schedule where each node competes for channel access when it needs to transmit data without any guarantee of success transmission. Hence, a hybrid MAC protocol that adopts Time Division Multiple Access (TDMA) and Carrier Sense Multiple Access (CSMA/CA) is required to enhance the network performance. Nevertheless, hybrid MAC protocols cannot ensure services differentiation so that critical messages and nonsecurity messages have the same priority. In this context and to overcome such limitation, this paper focuses on the design of a novel access method integrating TDMA and Enhanced Distributed Channel Access (EDCA) in unique architecture called Qos-aware Centralized Hybrid MAC protocol (QCH-MAC). To prove the efficiency of our proposal, the latter was implemented in ns-3 and the simulator SUMO to generate vehicular mobility traces. Therefore, experiments were performed to show the high performance of our system

International audience; Satellite image time-series (SITS) methods have contributed notably to detection of global change over the last decades, for instance by tracking vegetation changes. Compared with multi-temporal change detection methods, temporally highly resolved SITS methods provide more information in a single analysis, for instance on the type and consistency of change. In particular, SITS decomposition methods show a great potential in extracting various components from non-stationary time series, which allows for an improved interpretation of the temporal variability. Even though many case studies have applied SITS decomposition methods, a systematic comparison of common algorithms is still missing. In this study, the seasonal trend loess (STL), breaks for additive season and trend (BFAST) and multi-resolution analysis-wavelet transform (MRA-WT) were explored in order to evaluate their performance in modelling, monitoring and detecting land-cover changes with pronounced seasonal variations from simulated normal difference vegetation index time series. The selected methods have all proven their ability to characterize the non-stationary vegetation dynamics along with different physical processes driving the vegetation dynamics. Our results indicated that BFAST is the most accurate method for the examined simulated dataset in terms of RMSE, whereas MRA-WT showed a great potential for the extraction of multi-level vegetation dynamics. Considering the computational efficiency, both STL and MRA-WT outperformed BFAST.