Search

In this paper, we propose a method for coarse camera pose computation which is robust to viewing conditions and does not require a detailed model of the scene. This method meets the growing need of easy deployment of robotics or augmented reality applications in any environments, especially those for which no accurate 3D model nor huge amount of ground truth data are available. It exploits the ability of deep learning techniques to reliably detect objects regardless of viewing conditions. Previous works have also shown that abstracting the geometry of a scene of objects by an ellipsoid cloud allows to compute the camera pose accurately enough for various application needs. Though promising, these approaches use the ellipses fitted to the detection bounding boxes as an approximation of the imaged objects. In this paper, we go one step further and propose a learning-based method which detects improved elliptic approximations of objects which are coherent with the 3D ellipsoid in terms of perspective projection. Experiments prove that the accuracy of the computed pose significantly increases thanks to our method and is more robust to the variability of the boundaries of the detection boxes. This is achieved with very little effort in terms of training data acquisition -- a few hundred calibrated images of which only three need manual object annotation. Code and models are released at https://github.com/zinsmatt/3D-Aware-Ellipses-for-Visual-Localization., Comment: Presented at 3DV 2020. Code and models released at https://github.com/zinsmatt/3D-Aware-Ellipses-for-Visual-Localization

International audience; In the absence of a music score, tempo can only be defined by its perception by users. Thus recent studies have focused on the estimation of perceptual tempo defined by listening experiments. So far, algorithms have only been proposed to estimate the tempo when people agree on it. In this paper, we study the case when people disagree on the perception of tempo and propose an algorithm to predict this disagreement. For this, we hypothesize that the perception of tempo is correlated to a set of variations of various viewpoints on the audio content: energy, harmony, spectral-balance variations and short-term-similarity-rate. We suppose that when those variations are coherent, a shared perception of tempo is favoured and when they are not, people may perceive different tempi. We then propose several statistical models to predict the agreement or disagreement in the perception of tempo from these audio features. Finally, we evaluate the models using a test-set resulting from the perceptual experiment performed at Last-FM in 2011.

TEC - Travaux en Cours; International audience; Multi-display environments (MDE) are becoming more and more common in public areas. They are mainly used to display information due to the limited interaction possibilities they offer. In this paper, we study how using everyday objects, available on or around users, could improve interaction with public multi-display environments. To this end, we conducted a study, in the form of a creativity session, to understand how participants would use everyday objects, to realize the most usual tasks in public MDEs. The results of this study are presented and discussed in this paper.; Les environnements multi-écrans (EME) se multiplient dans les lieux publics. Les possibilités d’interaction qu’ils offrent étant limitées, ils sont principalement utilisés pour afficher de l’information. Dans cet article, nous étudions comment les objets du quotidien, disponibles sur ou autour des utilisateurs, pourraient porter l’interaction avec les environnements multi-écrans publics. Cette approche fait émerger des solutions pour des interactions rapides, opportunistes et naturelles, ce qui répond aux besoins liés à ce type d’environnements (publics et peu sécurisés). Nous avons réalisé une séance de créativité à laquelle ont participé 8 personnes. Le but de cette séance était d’étudier la manière avec laquelle les participants exploiteraient différentes catégories d'objets du quotidien pour réaliser plusieurs tâches d’interaction avec des EME.

Given the evolution of wireless technologies, Wireless Mesh Networks (WMN) have appeared as an emerging low-cost solution to ensure last-mile connectivity to the Internet network. However, providing real-time and streaming applications, such as VoIP (Voice over IP) and VoD (Video on Demand), with a satisfying QoS level is considered as an important challenge within these networks. In this paper, we propose a QoS based routing protocol, called Hybrid QoS Mesh Routing (HQMR), jointly with a clustering algorithm to improve the scalability of mesh networks. HQMR is composed of two routing sub- protocols: a reactive QoS based routing protocol for intra-mesh infrastructure communications and a proactive QoS based multi- tree routing protocol for communications with external networks. We analyze in this paper the simulation results of different scenarios conducted on the network simulator ns-3 to demonstrate the effectiveness of the reactive routing sub-protocol while forwarding real-time applications with a QoS guarantee.

International audience; Given the maturity of Internet standards, inter-organizational workflow is expected to be deployed in environments that are more dynamic and open than before. This paper addresses Inter Organizational Workflow (IOW) coordination in such a context, mainly investigating autonomic coordination managed at run-time. It is based on the idea that an agent-based approach is suitable to deal with this issue. More precisely, this paper introduces a framework for dynamic IOW in which involved processes are encapsulated into agents, called Process Agents (PA) in order to give them the capability to autonomously decide with whom, when and how to cooperate, and in which involved processes can access protocol components for their coordination needs. Our approach is based on the capability of PAs in playing different coordination protocols in order to take part in new business opportunities. This solution has numerous advantages. First, it provides extendable and reusable coordination components. Then, it supports run-time protocol integration. Finally, it eases openness since it imposes very few constraints.

International audience; Create an electronic dictionary from scratch is an expensive job because this task mobilizes over a long period, the work of skilled contributors, if not in lexicology, at least in linguistics. The use of specialized computer tools is essential for resources used by programs in natural language processing. When the socio-economic environment does not gather the necessary resources to the drafting of an electronic dictionary and printed dictionaries exist, these dictionaries are an important resource that can be used to initialize the creation of electronic lexical resources. This paper presents theoretical and practical aspects concerning the conversion of publishing dictionaries to electronic lexical resources. It takes into account the issue of limited economic resources, technology and the availability of qualified persons. Our field experiments concerns under-resourced languages mainly in Southeast Asia (Khmer, Malay, Vietnamese) and the Sahel (Bambara, Hausa, Kanuri, Tamajaq, Zarma), as most of the examples and socio-linguistic situations described in the paper relate to these areas. After a brief history devoted to the formats of electronic dictionaries (SGML, XML, XSLT and CSS), we present two standards that are dedicated to them (Text Encoding Initiative and Lexical Markup Framework). The issue of under-resourced languages is exposed and is followed by some examples concerning published dictionaries. The main technical challenges are detailed like the lack of standardization of the alphabets used and special characters (outside the traditional latin range). The conversion methodology is outlined and then detailed. The conversion to a bridge format in XML can be done by regular expressions or using specialized tools. Then, the bridge format is converted into the target format in LMF. The last part is dedicated to the consultation of resources through an online platform resource management.

International audience; In the next decades, the growth in population ageing will cause important problems to most industrialized countries. To tackle this issue, Ambient Assistive Living (AAL) systems can reinforce the well-being of elderly people, by providing emergency, autonomy enhancement, and comfort services. These services will postpone the need of a medicalized environment, and will allow the elderly to stay longer at home. However, each elderly has specific needs and a deployment environment of such services is likely unique. Furthermore, the needs evolve over time, and so does the deployment environment of the system. In this paper, we propose the use of a model-based development method, the adaptive medium approach, to enable dynamic adaptation of AAL systems. We also propose improvements to make it more suited to the AAL domain, such as considering heterogeneity and a composition model. The paper includes an evaluation of the prototype implementing the approach, and a comparison with related work.

This paper presents our recent work on multimedia fingerprinting, also known as traitor tracing. We focus on deriving a better accusation process for the well known Tardos codes. It appears that Tardos orginal decoding is very conservative: its performances are guaranteed whatever the collusion strategy. Indeed, major improvements stem from the knowledge of the collusion strategy. Therefore, this paper investigates how it is possible to learn and adapt to the collusion strategy. Our solution is based on an iterative algorithm a la EM, where a better estimation of the collusion strategy yields a better tracing of the colluders, which in return yields a better estimation of the collusion strategy etc.

Recent advances in telecommunication and positioning systems, web and wireless architectures offer new perspectives and challenges to the management and visualization of mobile and geo-referenced objects. Amongst many research challenges still opened, the development of integrated sensorbased architectures necessary to the integration of mobile data has been often neglected. Real-time integration of sensor-based data, their management within a distributed system, and diffusion to different levels of services and interfaces to what should be considered as a collaborative web GIS constitutes the objective of the research presented in this paper. This paper introduces a modular and experimental web GIS framework applied to maritime navigation, and where mobile objects behave in a maritime environment. Different levels of services are developed, including a web-based wireless access and interface to a traffic monitoring application.

International audience; System-on-Chip Methodologies & Design Languages brings together a selection of the best papers from three international electronic design language conferences in 2000. The conferences are the Hardware Description Language Conference and Exhibition (HDLCon), held in the Silicon Valley area of USA; the Forum on Design Languages (FDL), held in Europe; and the Asia Pacific Chip Design Language (APChDL) Conference. The papers cover a range of topics, including design methods, specification and modeling languages, tool issues, formal verification, simulation and synthesis. The results presented in these papers will help researchers and practicing engineers keep abreast of developments in this rapidly evolving field

The placement of Virtual Network Function - Forwarding Graphs (VNF-FGs) is one of the basic operations in the networks of the future. Being NP-hard, several heuristics and metaheuristics have been proposed. However, these approaches are inefficient due to the need to recalculate the solution at each service placement. In this paper, we adapt one of the most advanced approaches in Deep Reinforcement Learning (DRL), in order to improve exploration by generalizing the neural network calculating action values. We also propose an evolutionary algorithm to evolve these neural networks in order to discover better ones, which also avoids getting stuck in local minima. In order to avoid going through the almost innumerable number of infeasible solutions, we propose a heuristic, which combined with our DRL, makes it possible to guarantee the feasibility of the solutions and therefore to make the placement much more efficient. The simulation results we obtained confirm the quality of the solutions obtained as well as the superiority of the proposed solution over the existing one.

Renewable resources like fish stock or forests should be exploited at a rate that supports regeneration and sustainability—a complex problem that requires adaptive approaches to maintain a sufficiently high exploitation while avoiding depletion. In the presence of oblivious agents that cannot keep track of all available resources—a frequent condition in swarm robotics—ensuring that the exploitation effort is correctly balanced is particularly challenging. Additionally, the possibility to exploit resources by multiple robots opens the way to focusing the effort either on a single or on multiple resources in parallel. This means that the swarm needs to collectively decide whether to remain cohesive or split among multiple resources, as a function of the ability of the available resources to replenish after exploitation. In this paper, we propose a decentralised strategy for a swarm of robots that adapts to the available resources and balances the effort among them, hence allowing to maximise the exploitation rate while avoiding to completely deplete the resources. A detailed analysis of the strategy parameters provides insights into the working principles and expected performance of the robot swarm., SCOPUS: ar.j, info:eu-repo/semantics/published

This paper investigates the automatic monitoring of tool usage during a surgery, with potential applications in report generation, surgical training and real-time decision support. Two surgeries are considered: cataract surgery, the most common surgical procedure, and cholecystectomy, one of the most common digestive surgeries. Tool usage is monitored in videos recorded either through a microscope (cataract surgery) or an endoscope (cholecystectomy). Following state-of-the-art video analysis solutions, each frame of the video is analyzed by convolutional neural networks (CNNs) whose outputs are fed to recurrent neural networks (RNNs) in order to take temporal relationships between events into account. Novelty lies in the way those CNNs and RNNs are trained. Computational complexity prevents the end-to-end training of "CNN+RNN" systems. Therefore, CNNs are usually trained first, independently from the RNNs. This approach is clearly suboptimal for surgical tool analysis: many tools are very similar to one another, but they can generally be differentiated based on past events. CNNs should be trained to extract the most useful visual features in combination with the temporal context. A novel boosting strategy is proposed to achieve this goal: the CNN and RNN parts of the system are simultaneously enriched by progressively adding weak classifiers (either CNNs or RNNs) trained to improve the overall classification accuracy. Experiments were performed in a dataset of 50 cataract surgery videos and a dataset of 80 cholecystectomy videos. Very good classification performance are achieved in both datasets: tool usage could be labeled with an average area under the ROC curve of $A_z = 0.9961$ and $A_z = 0.9939$, respectively, in offline mode (using past, present and future information), and $A_z = 0.9957$ and $A_z = 0.9936$, respectively, in online mode (using past and present information only)., Accepted for publication in Medical Image Analysis

International audience; Success of Wikipedia is decidedly due to the free availability of high quality articles across many different expertise areas. If most of these resolute collaborations between authoritative users might constitute referenceable sources, Wikipedia is not sheltered from well-identified problems regarding articles quality, e.g., reputability of third-party sources and vandalism. Because of the huge number of articles and the intensive edit rate, it is not reasonable to even consider the manual evaluation of the content quality of each article. In this paper, we tackle the problem of modeling and predicting the quality of articles in collaborative platforms. We propose a quality model integrating both temporal and structural features captured from the implicit peer review process enabled by Wikipedia. A generic HITS-like framework is developed and able to capture both the quality of the content and the authority of the associated authors. Notably, a mutual reinforcement principle held between articles quality and author’s authority is exploited in order to take advantage of the collaborative graph generated by the users. Experiments conducted on a set of representative data from Wikipedia show the effectiveness of the computed indicators both in an unsupervised and supervised scenario.

International audience; Risks factor leading to type 2 diabetes and cardiovascular diseases can be reduced by a preventive care infrastructure able to modify the habits toward healthier lifestyles, including more exercise, a better diet and a reduced stress. Accessing the home space is important not only to monitor indoor environment quality (IEQ) but also to provide relevant user's feedback and engage users towards a healthy life style. This paper present a new preventive care infrastructure composed of a distributed infrastructure for heterogeneous devices in the home local area network (named xAAL) in conjunction with a machine to machine protocol (MQTT) to external health & well-being services. Finally, end-user tests were performed in the Experiment'HAAL living lab in order to test sensors and feedbacks.

International audience; Genetic Regulatory Networks (GRNs) implementations have a high degree of variability in their details. Parameters, encoding methods, and dynamics formulas all differ in the literature, and some GRN implementations have a high degree of model complexity. In this paper, we present a comparative study of different implementations of a GRN and introduce new variants for comparison. We use a modified Genetic Algorithm (GA) to evaluate GRN performance on a number of common benchmark tasks, with a focus on real-time control problems. We propose an encoding scheme and set of dynamics equations that simplifies implementation and evaluate the evolutionary fitness of this proposed method. Lastly, we use the comparative modifications study to demonstrate overall enhancements for GRN models.

International audience; Operating systems support autoscroll to allow users to scroll a window while in dragging mode: the user moves the pointer near the window’s edge to trigger an "automatic" scrolling. Scrolling rate is typically proportional to the distance between the pointer and the window’s edge. This approach suffers from several problems, especially when the window is maximized resulting in a very limited space around the window. An other problem is that for some operations, such as object drag-and-drop, the source and destination might be located in different windows, making it complicated for the computer system to understand user’s intention. In this paper, we present ForceEdge, a novel autoscroll technique relying on trackpads with force-sensing capabilities to alleviate all the problems related to autoscroll. We present the theoretical foundations of ForceEdge and the implementation of a demonstrator that can be used to compare ForceEdge to the other autoscroll methods.; Les systèmes d’exploitation supportent le défilement automatique (autoscroll) dans le but de permettre aux utilisateurs de faire défiler une fenêtre depuis le mode glisser: l’utilisateur déplace le pointeur de la souris vers le bord de la vue, ce qui a pour effet de commencer à faire défiler cette dernière "automatiquement". La vitesse de défilement est généralement associée à la distance entre le pointeur de la souris et le bord de la fenêtre. Cette approche souffre de plusieurs problèmes, notamment que l’espace autour de la fenêtre peut être limité, voire inexistant, par exemple quand une vue est maximisée. Un autre problème est que pour certaines opérations, comme le glisser-déposer d’objets, il est possible que la destination se trouve dans une autre vue que celle d’origine, ce qui crée une situation potentiellement ambiguë pour le système. Dans cet article, nous présentons ForceEdge, une nouvelle technique d’autoscroll qui tire parti des capacités des pavés tactiles récents pour s’affranchir des problèmes inhérents aux méthodes conventionnelles d’autoscroll. Nous présentons les fondements théoriques de ForceEdge et l’implémentation d’un démonstrateur qui permet de comparer ForceEdge aux techniques d’autoscroll par défaut de Mac OS X.

International audience; In this paper, we show how a dangerousness metric can be used to modify the input of a gene regulatory network when plugged to a virtual car. In the context of the 2015 Simulated Car Racing Championship organized during GECCO 2015, we have developed a new cartography methodology able to inform the controller of the car about the incoming complexity of the track: turns (slipperiness, angle, etc.) and bumps. We show how this dangerousness metric improves the results of our controller and outperforms other approaches on the tracks used in the competition.

International audience; Knowledge-based decision support systems (KBDSS) have evolved greatly over the last few decades. The key technologies underpinning the development of KBDSS can be classified into three categories: technologies for knowledge modelling and representation, technologies for reasoning and inference and web-based technologies. In the meantime, service systems have emerged and become increasingly important to value adding activities in the current knowledge economy. This paper provides a review on the recent advances in the three types of technologies, as well as the main application domains of KBDSS as service systems. Based on the examination of literature, future research directions are recommended for the development of KBDSS in general and in particular to support decision-making in service industry.

International audience; Ordering events in a distributed systems consists fundamentally in delaying event delivery. Partial ordering, such as FIFO and causal order, has many practical usages in distributed systems and can be obtained in arbitrarily large and dynamic networks. However, partial orderings imply that messages cannot be sent and delivered as soon as produced. In this paper, we examine the latency induced by such partial orderings. We obtain a probabilistic measure of the moment a message can be delivered according the different characteristics of the distributed system. Having such a measure helps to understand the systems behavior and to design new protocols. For instance, our measure allows us to parametrize a naive, albeit efficient, fault-tolerant causal delivery mechanism. We experimentally validate our approach using Internet-scale production distribution latency including faults.

TAAABLE is a Case-Based Reasoning (CBR) system that uses a recipe book as a case base to answer cooking queries. TAAABLE participates in the Computer Cooking Contest since 2008, and succeeds thanks to a smart combination of various knowledge-based system methods and techniques, such as: CBR, knowledge representation, acquisition, discovery and management, and natural language processing. In this paper, we first present the CBR engine and its features such as the retrieval process based on minimal generalization of a query and different adaptation processes. Then the knowledge containers used by the system, and how they were acquired is detailed. To conclude, we discuss some research developments resulting from this application.

International audience; With the emergence of reconfigurable FPGA circuits as a credible alternative to GPUs for HPC acceleration, new compilation paradigms are required to map high-level algorithmic descriptions to a circuit configuration (High-Level Synthesis, HLS). In particular, novel parallelization algorithms and intermediate representations are required. In this paper, we present the data-aware process networks (DPN), a dataflow intermediate representation suitable for HLS in the context of high-performance computing. DPN combines the benefits of a low-level dataflow representation-close to the final circuit-and affine iteration space tiling to explore the parallelization trade-offs (local memory size, communication volume, parallelization degree). We outline our compilation algorithms to map a C program to a DPN (front-end), then to map a DPN to an FPGA configuration (back-end). Finally, we present synthesis results on compute-intensive kernels from the Polybench suite.

International audience; For over a decade, MapReduce has become a prominent programming model to handle vast amounts of raw data in large scale systems. This model ensures scalability, reliability and availability aspects with reasonable query processing time. However these large scale systems still face some challenges: data skew, task imbalance, high disk I/O and redistribution costs can have disastrous effects on performance. In this paper, we introduce MRFA-Join algorithm: a new frequency adaptive algorithm based on MapReduce programming model and a randomised key redistribution approach for join processing of large-scale datasets. A cost analysis of this algorithm shows that our approach is insensitive to data skew and ensures perfect balancing properties during all stages of join computation. These performances have been confirmed by a series of experimentations.

International audience; ROV 3D project aims at developing innovative tools which link underwater photogrammetry and acoustic measurements from an active underwater sensor. The results will be 3D high resolution surveys of underwater sites and landscapes useful to keep in memory cultural and natural heritage. The new means and methods developed aim at reducing the investigation time in situ, and proposing comprehensive and non-intrusive measurement tools for the studied environment. In this paper, we are presenting a new method of 3D surveys which are dedicated to high resolution modeling of underwater sites. The main met constraints in situ are taken into account and this method leads to a precise 3D reconstruction. Some examples will present both the main obtained results and their limitations. We will end with the perspectives and the necessary improvements to the method, so as to automate the multimodal registration step.

National audience; Dans cet article, nous nous sommes intéressés aux problèmes de reconnaissance noncoopérative de cibles (NCTR) en tant que problème de classification supervisée. Après une présentation du système d'acquisition des profils distance radar et du problème de reconnaissance, suivie d'une étude statistique des données, nous proposons d'utiliser un algorithme des K plus proches voisins (KPPV) dont les performances sont détaillées en fonction du nombre de voisins K, du type de distance utilisée et de la nature des données utilisées (débruitées ou non). Dans un second temps, cet algorithme a été parallélisé sur un processeur many-coeurs (GPU : Graphics Processing Unit). Les opérations arithmétiques et le modèle d'accès mémoire ont été étudiés pour obtenir la meilleure parallélisation des calculs. Enfin, nous terminons par une discussion autour des perspectives envisageables pour la méthode proposée, notamment en s'intéressant à d'autres espaces de représentation ou à d'autres méthodes de classification. ABSTRACT. In this paper, first, we present the problem of Non Cooperative Target Recognition (NCTR) as a supervised classification problem. After a presentation on the radar acquisition system of range profiles and the problem of recognition, followed by a statistical study of data, we use a classical classification method of K Nearest Neighbors (KNN) to do this classification. We explore and compare the performances of this algorithm based on the choice of the distances, the choice of K and the nature of used data (denoised or not). KNN algorithm has been executed initially on CPU with Matlab and then on GPU. Arithmetic operations and memory access pattern has been studied to get the best parallelization. Finally, we conclude with a discussion about possible perspectives for the proposed method especially by focusing on other representation spaces or other classification methods.

International audience; Collection and distribution of Radio Frequency IDentification (RFID) data are subject to various privacy concerns. These concerns are of paramount importance when sensitive data are processed (e.g., medical data). Therefore, it is crucial to treat sensitive data privacy in early stages to master the data view for upper layers and to minimize, as soon as possible, the risk of unauthorized disclosures. While most recent works focus on securing the access and visibility of collected information in the final databases, data processed in the middleware do not seem involved in the process of privacy protection. Current EPCglobal standards for RFID also suffer from insufficient attention to this issue. In this paper, we propose a privacy controller module that enhances the Filtering and Collection (F&C) middleware of the EPCglobal network. We provide a privacy policy-driven model, using some enhanced contextual concepts of the extended Role Based Access Control model. The feasibility of our privacyenhanced model is shown by integrating our solution into the F&C middleware of the Fosstrak framework, an open-source implementation of the EPCglobal network specifications.

There is a rise in the number 3D audio-visual productions and archives that creates a need for indexation of 3D contents. Event detection using audio modality is a difficult task. The standard way to do classification on 3D audio is to first down-mix to mono audio and classify on that. In this paper, we describe a generic classifier for multi-channel audio event detection and propose several information fusion strategies. Our system is evaluated on a speech and music detection task on the audio of 3D movies. We improve the classification performances on our database by 1.5% for speech detection, and 8% for music detection, compared to the standard downmixing method. We also provide a comparison of several information fusion methods in the experiments.

National audience; Recently government agencies have started to utilize innovative contracting methods that provide incentives for improving construction quality. These emerging contracting methods place an enormous pressure on the contractors to improve construction quality. For a general contractor, which subcontracts most tasks of a project and invites a number of bids, choosing an appropriate bid which satisfies the time, cost and quality of construction project is complex and challenging. To solve this problem involving conflicting objectives, a fuzzy clustering-based genetic algorithm (FCGA) approach is proposed in this paper. A case study of highway construction is used to demonstrate the applicability of the proposed approach. A comparative study is conducted over three test cases involving varying dimensions and complexities to test performance of the proposed FCGA against existing approaches. Results reveal that the FCGA is capable of generating better Pareto front than other existing approaches.

This paper presents the latest improvements on AudioPrint: the IRCAM audio fingerprint system. Cosine filters are introduced in the short-term spectral analysis, in order to compensate the effect of pitch shifting, and a simple solution is proposed for the determination of the frame positions, robust to audio degradations, with nearly no additional cost. We then show that both contributions significantly improve the Audio-Print system, with evaluations both on a free corpus, made publicly available, and a real-world corpus of broadcast radio streams.

In this paper, we introduce a new approach to music structure segmentation that is based on the joint estimation of structural segments, keys and chords in one probabilistic framework. More precisely, the boundaries of a structure segment are determined by detecting key changes and by utilizing the difference in prior probability of chord transitions according to their position in a structural segment. In contrast to many of the recent approaches to structural segmentation, this system does not work with self-similarity matrices, although it has been designed to integrate this kind of approach into the framework at a later stage. However, just the current version of the system, using only the estimated harmony, is already producing encouraging results, especially with respect to the precise localization of the boundaries.

International audience; Metamorphic robots are robots that can change their shape by reorganizing the connectivity of their modules to adapt to new environments, perform new tasks, or recover from damages. In this paper we present a decentralized method for structural evolving of a class of lattice-based simulated metamorphic robots in a static environment. These robots are considered as a set of crystalline (compressible) modules that are able to connect or disconnect one from each another or even exchange information and energy with the neighbor modules in order to form various structures/patterns dynamically. Our approach is splitted in two layers: in the first layer a genetic algorithm is used to generate a number of well suited target configurations based on current information perceived from environment, while in the second layer a PacMan-like algorithm is used to make a plan for modules movement to transform the robot from its current pattern to the target pattern emerged in first layer.

International audience; This paper describes the first step of a research project with the aim of predicting students' performance during an online curriculum on a LMS and keeping them from falling behind. Our research project aims to use data mining, machine learning and artificial intelligence methods for monitoring students in e-learning trainings. This project takes the shape of a partnership between computer science / artificial intelligence researchers and an IT firm specialized in e-learning software. We wish to create a system that will gather and process all data related to a particular e-learning course. To make monitoring easier, we will provide reliable statistics, behaviour groups and predicted results as a basis for an intelligent virtual tutor using the mentioned methods. This system will be described in this article. In this step of the project, we are clustering students by mining Moodle log data. A first objective is to define relevant clustering features. We will describe and evaluate our proposal. A second objective is to determine if our students show different learning behaviours. We will experiment whether there is an overall ideal number of clusters and whether the clusters show mostly qualitative or quantitative differences. Experiments in clustering were carried out using real data obtained from various courses dispensed by a partner institute using a Moodle platform. We have compared several classic clustering algorithms on several group of students using our defined features and analysed the meaning of the clusters they produced.

National audience; In this paper we review approaches for evaluating information retrieval systems using test collections. We first give the definition of a test collection and present the main metrics used in literature to evaluate systems. We then show, thanks to three examples (search results clustering, automatic summarization and image retrieval), the variety of the existing evaluation frameworks.; L’objectif de cet article est de présenter un panorama de l’évaluation des systèmes de recherche d’information se basant sur des collections de référence. Nous détaillons dans un premier temps ce qu’est une collection de référence ainsi que les mesures d’évaluation associées. Nous développons ensuite les problématiques d’évaluation à travers trois cadres de recherche d’information spécifiques : le clustering de documents, le résumé automatique et la recherche d’images par l’exemple, et montrons la variété des mesures et collections de référence existantes.

International audience; In this paper, we highlight how wikis can be relevant solutions for building applications in domains characterized by a rapid evolution of knowledge. We will point out the capabilities of semantic extension to provide better quality of content, to improve searching, to support complex queries and finally to carry out on different type of users. We detail the architecture of WikiBridge, a semantic wiki, its annotation model and the consistency checking mechanism. A specific section is dedicated to the description of the CARE project and ontology design which is the compulsory foundational knowledge for the application.

National audience; Supplier selection is nowadays one of the critical topics in supply chain management. This paper presents a new decision making approach for group multi-criteria supplier selection problem, which clubs supplier selection process with order allocation for dynamic supply chains to cope market variations. More specifically, the developed approach imitates the knowledge acquisition and manipulation in a manner similar to the decision makers who have gathered considerable knowledge and expertise in procurement domain. Nevertheless, under many conditions, exact data are inadequate to model real-life situation and fuzzy logic can be incorporated to handle the vagueness of the decision makers. As per this concept, fuzzy-AHP method is used first for supplier selection through four classes (CLASS I: Performance strategy, CLASS II: Quality of service, CLASS III: Innovation and CLASS IV: Risk), which are qualitatively meaningful. Thereafter, using simulation based fuzzy TOPSIS technique, the criteria application is quantitatively evaluated for order allocation among the selected suppliers. As a result, the approach generates decision-making knowledge, and thereafter, the developed combination of rules order allocation can easily be interpreted, adopted and at the same time if necessary, modified by decision makers. To demonstrate the applicability of the proposed approach, an illustrative example is presented and the results analyzed.

International audience; Internet connectivity in the ITS context is a flourishing demand that has to be covered by efficient information and communication technologies. Thanks to the provision of this connectivity, not only end users will profit from communication services, but also services for controlling the operation of the vehicle will benefit from the Internet connectivity. Services related to the operation and control of the vehicles exchange sensitive data, so strong security properties have to be ensured for this type of services. In addition, because of the privacy concerns related to the end users, it is desired to provide this kind of services with an adequate security level. From the communication point of view, a vehicle can be regarded as a mobile network where nodes onboard obtain continuous and optimal Internet connectivity, so, its mobility has to be managed. In the same way, network mobility management protocols should not only avoid security leaks, but also they have to ensure an adequate security level. It is needless to say that the introduction of security properties cannot render not fulfilling service performance requirements. In fact, the trade-off between security and application performance is a must. ITS standardisation bodies have adopted the NEtwork MObility Basic Support (NEMO BS) protocol to manage the mobility of networks. However, it still presents shortcomings like lack of security support and routing problems, which leads to a bad performance. One of the most promising design approaches is to consider a base mobility management protocol that provides out of the box security and route optimisation support like the Host Identity Protocol (HIP). Different solutions based on HIP can be found to solve network mobility in the literature, but none of these solutions aim at solving securely and efficiently network mobility management. That is, the provision of security properties to the network mobility management itself and to the end-to-end data communications while not increasing the signalling overhead and the manageability level remains unsolved. In this paper, we present the NeMHIP. NeMHIP is a secure and efficient network mobility management protocol which is based on HIP. In order to demonstrate its feasibility, we have carried out a study by means of analytical modelling to assess the performance of user applications with stringent QoS requirements like VoIP. Results demonstrate that the introduction of the NeMHIP in the ITS context is feasible because security properties are ensured while application performance requirements are satisfied. Therefore, we successfully achieve the trade-off between security and performance.

MusiClef is a multimodal music benchmarking initiative that will be running a MediaEval 2012 Brave New Task on Multimodal Music Tagging. This paper describes the setup of this task, showing how it complements existing benchmarking initiatives and fosters less explored methodological directions in Music Information Retrieval. MusiClef deals with a concrete use case, encourages multimodal approaches based on these, and strives for transparency of results as much as possible. Transparency is encouraged at several levels and stages, from the feature extraction procedure up to the evaluation phase, in which a dedicated categorization of ground truth tags will be used to deepen the understanding of the relation between the proposed approaches and experimental results.

National audience; In this paper, we present a general framework for supervised classification. This framework only needs the definition of a generalisation operator and provides ensemble methods. For sequence classification tasks, we show that grammatical inference has already defined such learners for automata classes like reversible automata or k-TSS automata. Then we propose a generalisation operator for the class of balls of words. Finally, we show through experiments that our method efficiently resolves sequence classification tasks.; Dans cet article, nous présentons un cadre d'apprentissage général pour la classification supervisée. Ce cadre ne nécessite que la définition d'un opérateur de généralisation et fournit en particulier des méthodes d'ensemble. Pour les tâches de classification de séquences, nous montrons que l'inférence grammaticale, avec des objectifs différents, a déjà défini de tels apprenants pour certaines familles d'automates comme les réversibles ou les k-TSS. Nous proposons ensuite un opérateur de généralisation original pour la famille des boules de mots. Enfin, nous montrons au travers de différentes expérimentations que notre approche permet effectivement de résoudre des tâches de classification de séquences.

LGEP 2011 ID = 739; International audience; This paper shows that Ridge Polynomial Neural Networks (RPNN) and Least-Square Support Vector Machines (LS-SVM) technique provide efficient tools for microwave characterization of dielectric materials. Such methods avoids the slow learning properties of multilayer perceptrons (MLP) which utilize computationally intensive training algorithms and can get trapped in local minima. RPNN and LS-SVM are combined in this work with the Finite Element Method (FEM) to evaluate the dielectric materials properties. The RPNN is constructed from a number of increasing orders of Pi-Sigma units, it maintains fast learning properties and powerful mapping capabilities of single layer High Order Neural Networks (HONN). LS-SVM is a statistical learning method that has good generalization capability and learning performance. The FEM is used to create the data set required to train the RPNN and LS-SVM. The performance of a LS-SVM model depends on a careful setting of its associated hyper-parameters. In this study the LS-SVM hyper-parameters are optimized by using a Bayesian regularization technique. Results show that LS-SVM can achieve good accuracy and faster speed than neural network methods.

When tackling real-life datasets, it is common to face the existence of missing values within data. Explaining the origin of the missing values appearance allows to better control the quality of the data, as well as proposing suitable handling methods, e.g., their completion. The abundant literature heavily relies on the missing value appearance models proposed by Little and Rubin. However, a careful scrutiny of these statistic-based models highlights that they constitute an actual hamper towards their use by data mining techniques. The main thrust of this paper is the proposition of a new model for missing values appearance. Such introduced models rely on the use of the proper implication basis. MOTS-CLES: fouille de donnees, valours manquantes, donnees incompletes, regles d'association.

In this paper, we introduce a new partial consistency for constraint networks which is called Structural Consistency of level w and is denoted w-SC consistency. This consistency is based on a new approach. While conventional consistencies generally rely on local properties extended to the entire network, this new partial consistency considers global consistency on subproblems. These subproblems are defined by partial constraint graphs whose tree-width is bounded by a constant w. We introduce a filtering algorithm which achieves w-SC consistency. We also analyze w-SC filtering w.r.t. other classical local consistencies to show that this consistency is generally incomparable although this consistency can be regarded as a special case of inverse consistency. Finally, we present experimental results to assess the usefulness of this approach. We show that w-SC is a significantly more powerful level of filtering and more effective w.r.t. the runtime than SAC. We also show that w-SC is a complementary approach to AC or SAC. So we can offer a combination of filterings, whose power is greater than w-SC or SAC.

Extensive research in recent years has shown the benefits of \textit{cognitive radio} technologies to improve the flexibility and efficiency of spectrum utilization. This new communication paradigm, however, requires a well-designed spectrum allocation mechanism. In this paper, we propose an auction framework for cognitive radio networks to allow unlicensed secondary users (SUs) to share the available spectrum of licensed primary users (PUs) fairly and efficiently, subject to the interference temperature constraint at each PU. To study the competition among SUs, we formulate a non-cooperative multiple-PU multiple-SU auction game and study the structure of the resulting equilibrium by solving a non-continuous two-dimensional optimization problem. A distributed algorithm is developed in which each SU updates its strategy based on local information to converge to the equilibrium. We then extend the proposed auction framework to the more challenging scenario with free spectrum bands. We develop an algorithm based on the no-regret learning to reach a correlated equilibrium of the auction game. The proposed algorithm, which can be implemented distributedly based on local observation, is especially suited in decentralized adaptive learning environments as cognitive radio networks. Finally, through numerical experiments, we demonstrate the effectiveness of the proposed auction framework in achieving high efficiency and fairness in spectrum allocation.

In this paper, we present the design and implementation of a new approach for anomaly detection and classification over high speed networks. This approach is efficient in terms of its scalability, the anomaly detection delays and its ability not only to detect an anomaly but also to provide detailed and accurate information about the ongoing attack (attack type, IP addresses, port numbers, etc.). The proposed approach is based first of all on a data reduction phase through flow sampling by focusing mainly on short lived flows. The second step is then a random aggregation of some descriptors such as a number of SYN packets per flow in two different data structures called Count Min Sketch and Multi-Layer Reversible Sketch. A sequential change point detection algorithm continuously monitors the sketch cell values. An alarm is raised if a significant change is identified in cell values. We make profit of the reversibility properties of the Multi-Layer Reversible Sketch to retrieve useful information about which ''flow'' was the culprit one. The good properties of the Count Min Sketch in terms of resistance to collisions make it possible to check the accuracy of the information about the ongoing attack. With an appropriate definition of the combination of IP header fields that should be used to identify one flow we are able not only to detect the anomaly but also to classify the anomaly as DoS, DDoS or flash crowd, network scanning and port scanning. We validate our framework for anomaly detection on various real world traffic traces and demonstrate the accuracy of our approach on these real-life case studies. Some of our test traffic traces are measurements obtained on a geographical and technical subdivision of an ADSL network of a major Internet Service Provider. Other traces were collected in the framework of a French national research project during online experiments on a distributed measurement network with well known attack types and instants. Our analysis results from online implementation of our algorithm over measurements gathered by a DAG sniffing card are very attractive in terms of accuracy and response time. The proposed approach is very effective in detecting and classifying anomalies, and in providing information by extracting the culprit flows with a high level of accuracy.

National audience; In this paper we study the classical problem of the impact of an update on a view defined over semi-structured data. We adopt the following working hypotheses: (i) the source document is modeled by an unranked, labeled, ordered tree, (ii) a view V is a tree query whose evaluation on the source document provides a desired partial view of the document, (iii) a class of updates C. is also given by a tree query selecting the nodes to modify. We then study the following problem: given a view query V and a class of updates C., is it possible to detect if the view V is independent of each update q in C.? We show that the problem is in general PSPACE-hard. We propose a sufficient condition evaluable in polynomial time ensuring the independence of a view V with respect to a class of updates C.. We then consider the class of linear view queries for which the problem becomes polynomial. We also show that the tree query model chosen to express V and C., is incomparable with XPath but is able to capture positive queries of CoreXPath.

In 1997, B. Selman and H. Kautz proposed a series of 10 challenges. One of them concerned the design of a practical stochastic local search procedure for proving unsatisfiability (Challenge 5). Today, more than 10 years later, only few attempts were led to address this challenge, in spite of the great number of incomplete methods for proving satisfiability. In this paper, we propose a two steps algorithm for proving unsatisfiability of CNF formulas\footnote{supported by ANR UNLOC project n\ensuremath{^\circ} BLAN08-1\_328904}. The first step consists in detecting $\ssi$-gates greedily. At the same time, a polynomial algorithm is used to eventually prove the unsatisfiability of the extracted set of equivalency-gates. We show that this method outperforms the existing ones on some classes of instances. This method is then extended to other logical functions (and-gates and or-gates).

[Departement_IRSTEA]Ecotechnologies [TR1_IRSTEA]MOTIVE [Coll_IRSTEA]Series: Lecture Notes in Computer Science; National audience; Models for representing multidimensional systems usually consider that facts and dimensions are two different things. In this paper we propose a model based on UML which unifies the representations of fact and dimension member. Since a given element can play the role of a fact or a dimension member, this model permits more flexibility in the design and the implementation of multidimensional systems. Moreover this model offers the possibility to express various constraints to guarantee desirable properties for data. We then show that this model is able to handle most of the hierarchies which have been suggested to take real situations into account and to characterize certain properties of summarizability. Using this model we propose a complete development cycle of a multidimensional system. It appears that this cycle can be partially automated and that an end user can control the design and the implementation of his system himself.

The issue addressed in this paper is the semantics of simple vs complex (or synthetic vs analytic) expressions, in particular that of complex prepositions in Medieval French. We present the categorial structure of prepositions in Old and Middle French, and go on to show, on the basis of a corpus study of a few simple and complex prepositions, that the latter are indeed less polysemous than simple prepositions, as one could await, but nevertheless demonstrate a certain degree of polysemy. Their semantic evolution includes, for some, leaps accross conceptual domains - metaphorical transfers from space to time, for instance -, but it seems that it differs to a certain degree from that of simple prepositions, though this needs to be verified by further research., L'objectif de cet article est d'étudier le sémantisme des prépositions simples et complexes, en particulier en français médiéval. Nous présentons la structure de la catégorie des prépositions en français ancien et moyen, puis montrons, en partant de l'étude de quelques prépositions simples et complexes, que ces dernières sont en effet moins polysémiques que les prépositions simples, comme on peut s'y attendre, mais demeurent relativement polysémiques. Leur évolution sémantique inclut, pour certaines, des passages d'un domaine conceptuel à un autre : transferts métaphoriques de l'espace au temps, par exemple ; s'il semble bien que ces évolutions diffèrent de celles des prépositions simples, cela devra être analysé plus en détail dans des recherches futures.

International audience; This paper proposes a local, cooperative and real-time multi-agent approach to build adaptive and incremental profiles where a user is supposed to be represented by a set of textual documents. These documents are sequentially parsed, which leads to the construction of a temporary terminological network (TTN). This network is then merged with the TTN resulting from other documents, in order to create a permanent terminological network (PTN). This PTN is representative of the studied collection and can be used to index this collection thanks to a clustering approach. Preliminary results of the built system as well as perspectives are then presented.; Ce papier propose une approche multi-agent locale, coopérative et temps réel pour la création de profils adaptatifs et incrémentaux dans laquelle un usager est supposé être représenté par un ensemble de documents textuels. Ces documents sont analysés séquentiellement conduisant alors à la création d’un réseau terminologique temporaire (RTT). Ce réseau est ensuite fusionné avec les RTT issus d’autres documents, afin de générer un réseau terminologique permanent (RTP) représentatif de la collection étudiée et pouvant être ensuite indexé pour en extraire un profil. Les résultats préliminaires obtenus ainsi que les perspectives associées sont enfin présentés.

National audience; Our work focus on data management in Vehicular Ad Hoc Networks (VANETs). Many pieces of information may be exchanged in such networks, for instance to warn drivers when a potentially dangerous event arises (accident, emergency braking, obstacle in the road, etc.) or to try to assist them (available parking spaces, traffic congestions, real-time traffi conditions, etc.). Existing systems only use the data exchanged to warn the driver. Then, the data is considered obsolete and is deleted. In this paper, we rather propose to aggregate the data once it becomes obsolete. Our objective is to produce additional knowledge to be used by drivers when no relevant data has been communicated by neighbouring vehicles. For example, it becomes so possible to dynamically detect potentially dangerous road segments or to determine the areas where the probability to find an available parking space is high when none has been received

In this paper, a new evaluation of the complexity of forward checking for solving non-binary CSPs with finite domains is proposed. Unlike what is done usually, it does not consider the size of domains, but the size of the relations associated to the constraints. It may lead sometimes to define better complexity bounds. By using this first result, we show that the tractability hierarchy proposed in [6] which compares different methods based on a decomposition of constraint networks can be seen from a new viewpoint.