Publications

2019

• Discovery of usage patterns in digital library web logs using Markov modeling
  
  • Nouvellet Adrien
  • Beaudouin Valérie
  • d'Alché-Buc Florence
  • Prieur Christophe
  • Roueff François
  , 2019. This paper proposes a family of tools based on Markov modeling to quantitatively analyze how people access the digital collections of the Bibliothèque nationale de France (BnF, the national library of France), through the web platform called Gallica. The aim is to provide the BnF with relevant information about the various usage patterns to help them to better understand their users, improve the mediation efforts and the design of the website, in order to increase the general public use of the 4M-documents collection. For that purpose, the study focuses on the access logs retrieved from the Apache HTTP servers of Gallica that are converted into sequences of actions. In order to study user navigation behaviors, we propose to model the access log data using Markov Models, whether it be Markov chains when considering sequences of actions without duration, or Markov processes when taking into account duration. Our models are either used to capture an average behavior through meaningful statistics or to cluster the data to exhibit various interpretable types of usage. The numerical results bring new insights on the way the users interact with the platform, highlighting the mean duration of some actions such as the interaction with the search engine or the consultation of documents. Even if our approach requires the use of additional information in order to properly interpret the models and the correlations that it highlights, it is able to discover all types of behaviors, including the stealthiest and the most difficult to capture in traditional surveys, giving them their fair weight in terms of audience. We also show how this approach fits into a broader work combining data mining and ethnography.
• Multiple characters transforms and generalized Boolean functions
  
  • Mesnager Sihem
  • Riera Constanza
  • Stanica Pantelimon
  Cryptography and Communications - Discrete Structures, Boolean Functions and Sequences, Springer, 2019, 11 (6), pp.1247-1260. In this paper we investigate generalized Boolean functions whose spectrum is flat with respect to a set of Walsh-Hadamard transforms defined using various complex primitive roots of 1. We also study some differential properties of the generalized Boolean functions in even dimension defined in terms of these different characters. We show that those functions have similar properties to the vectorial bent functions. We next clarify the case of gbent functions in odd dimension. As a by-product of our proofs, more generally, we also provide several results about plateaued functions. Furthermore, we find characterizations of plateaued functions with respect to different characters in terms of second derivatives and fourth moments. (10.1007/s12095-019-00383-5)
  DOI : 10.1007/s12095-019-00383-5
• Interoperability Issues in Optical Transport Networks
  
  • Alahdab Luay
  • Ware Cédric
  • Le Rouzic Esther
  • Meuric Julien
  • Triki Ahmed
  , 2019 (We.B1.2). (10.1109/ICTON.2019.8840156)
  DOI : 10.1109/ICTON.2019.8840156
• A Survey of Neural Network Applications in Fiber Nonlinearity Mitigation
  
  • Catanese Clara
  • Triki Ahmed
  • Pincemin Erwan
  • Jaouën Yves
  , 2019.
• Arbitration-Induced Preemption Delays
  
  • Hebbache Farouk
  • Brandner Florian
  • Jan Mathieu
  • Pautet Laurent
  , 2019. The interactions among concurrent tasks pose a challenge in the design of real-time multi-core systems, where blocking delays that tasks may experience while accessing shared memory have to be taken into consideration. Various memory arbitration schemes have been devised that address these issues, by providing trade-offs between predictability, average-case performance, and analyzability. Time-Division Multiplexing (TDM) is a well-known arbitration scheme due to its simplicity and analyzability. However, it suffers from low resource utilization due to its non-work-conserving nature. We proposed in our recent work dynamic schemes based on TDM, showing work-conserving behavior in practice, while retaining the guarantees of TDM. These approaches have only been evaluated in a restricted setting. Their applicability in a preemptive setting appears problematic, since they may induce long memory blocking times depending on execution history. These blocking delays may induce significant jitter and consequently increase the tasks' response times. This work explores means to manage and, finally, bound these blocking delays. Three different schemes are explored and compared with regard to their analyzability, impact on response-time analysis, implementation complexity, and runtime behavior. Experiments show that the various approaches behave virtually identically at runtime. This allows to retain the approach combining low implementation complexity with analyzability. (10.4230/LIPIcs.ECRTS.2019.19)
  DOI : 10.4230/LIPIcs.ECRTS.2019.19
• Max-plus Operators Applied to Filter Selection and Model Pruning in Neural Networks
  
  • Zhang Yunxiang
  • Blusseau Samy
  • Velasco-Forero Santiago
  • Bloch Isabelle
  • Angulo Jesus
  , 2019. Following recent advances in morphological neural networks, we propose to study in more depth how Max-plus operators can be exploited to define morphological units and how they behave when incorporated in layers of conventional neural networks. Besides showing that they can be easily implemented with modern machine learning frameworks , we confirm and extend the observation that a Max-plus layer can be used to select important filters and reduce redundancy in its previous layer, without incurring performance loss. Experimental results demonstrate that the filter selection strategy enabled by a Max-plus is highly efficient and robust, through which we successfully performed model pruning on different neural network architectures. We also point out that there is a close connection between Maxout networks and our pruned Max-plus networks by comparing their respective characteristics. The code for reproducing our experiments is available online.
• Part-based approximations for morphological operators using asymmetric auto-encoders
  
  • Ponchon Bastien
  • Velasco-Forero Santiago
  • Blusseau Samy
  • Angulo Jesus
  • Bloch Isabelle
  , 2019. This paper addresses the issue of building a part-based representation of a dataset of images. More precisely, we look for a non-negative, sparse decomposition of the images on a reduced set of atoms, in order to unveil a morphological and interpretable structure of the data. Additionally, we want this decomposition to be computed online for any new sample that is not part of the initial dataset. Therefore, our solution relies on a sparse, non-negative auto-encoder where the encoder is deep (for accuracy) and the decoder shallow (for interpretability). This method compares favorably to the state-of-the-art online methods on two datasets (MNIST and Fashion MNIST), according to classical metrics and to a new one we introduce, based on the invariance of the representation to morphological dilation.
• Intégration des dimensions d’un véhicule dans un certificat d’authentification
  
  • Monteuuis Jean-Philippe
  • Zhang Jun J.
  • Labiod Houda
  • Servel Alain
  • Mafrica Stefano
  , 2019.
• Local Decoding and Update of Compressed Data
  
  • Vatedka Shashank
  • Tchamkerten Aslan
  , 2019, pp.572-576. In compressing large datasets it is often desirable to guarantee locality properties that allow the efficient decoding and efficient update of short fragments of data. This paper proposes a universal compression scheme for memoryless sources with the following features: 1. the rate can be made arbitrarily close to the entropy of the underlying source, 2. constant-sized (as a function of the blocklength) fragments of the source can be recovered by probing a constant number of codeword bits on average, 3. the update of constant-sized fragments of the source can be achieved by reading and modifying a constant number of codeword symbols on average, and 4. the overall encoding and decoding complexity is quasilinear in the blocklength of the source. (10.1109/ISIT.2019.8849634)
  DOI : 10.1109/ISIT.2019.8849634
• Exponent Trade-off for Hypothesis Testing Over Noisy Channels
  
  • Weinberger Nir
  • Kochman Yuval
  • Wigger Michèle
  , 2019, pp.1852-1856. The distributed hypothesis testing (DHT) problem is considered, in which the joint distribution of a pair of sequences present at separated terminals, is governed by one of two possible hypotheses. The decision needs to be made by one of the terminals (the "decoder"). The other terminal (the "encoder") uses a noisy channel in order to help the decoder with the decision. This problem can be seen as a generalization of the side-information variant of the DHT problem, where the rate-limited link is replaced by a noisy channel. A recent work by Salehkalaibar and Wigger has derived an achievable Stein exponent for this problem, by employing concepts from the DHT scheme of Shimokawa et al., and from unequal error protection coding for a single special message. In this work we extend the view to a trade-off between the two error exponents, additionally building on multiple codebooks and two special messages with unequal error protection. As a by product, we also present an achievable exponent trade-off for a rate-limited link, which generalizes Shimokawa et al.. (10.1109/isit.2019.8849485)
  DOI : 10.1109/isit.2019.8849485
• A Fundamental Storage-Communication Tradeoff in Distributed Computing with Straggling Nodes
  
  • Yan Qifa
  • Wigger Michele
  • Yang Sheng
  • Tang Xiaohu
  , 2019, pp.2803-2807. (10.1109/ISIT.2019.8849615)
  DOI : 10.1109/ISIT.2019.8849615
• Equality in the matrix entropy-power inequality and blind separation of real and complex sources
  
  • Rioul Olivier
  • Zamir Ram
  , 2019. The matrix version of the entropy-power inequality for real or complex coefficients and variables is proved using a transportation argument that easily settles the equality case. An application to blind source extraction is given. (10.1109/ISIT.2019.8849303)
  DOI : 10.1109/ISIT.2019.8849303
• An information-theoretic model for side-channel attacks in embedded hardware
  
  • Cherisey Eloi De
  • Guilley Sylvain
  • Rioul Olivier
  • Piantanida Pablo
  , 2019. Using information-theoretic tools, this paper estab- lishes a mathematical link between the probability of success of a side-channel attack and the minimum number of queries to reach a given success rate, valid for any possible distinguishing rule and with the best possible knowledge on the attacker’s side. This link is a lower bound on the number of queries, which depends on the mutual information between the traces and the secret key. This leads us to derive upper bounds on the mutual information that are as tight as possible and can be easily calculated. It turns out that, in the case of additive white Gaussian noise, the bound on the probability of success of any attack is directly related to the signal-to-noise ratio (SNR). This leads to easy computations and predictions of the success rate for any leakage model. (10.1109/ISIT.2019.8849763)
  DOI : 10.1109/ISIT.2019.8849763
• Détection erroné d’un faux à l’aide du système V2X
  
  • Monteuuis Jean-Philippe
  • Zhang Jun J.
  • Labiod Houda
  • Servel Alain
  • Mafrica Stefano
  , 2019.
• From Behavioural Contracts to Session Types
  
  • Fantechi Alessandro
  • Najm Elie
  • Stefani Jean-Bernard
  , 2019, pp.278-297. (10.1007/978-3-030-21485-2_16)
  DOI : 10.1007/978-3-030-21485-2_16
• Vers une exploitation efficace de grandes bases de connaissances par des graphes de contexte
  
  • Mimouni Nada
  • Moissinac Jean-Claude Jc
  , 2019, pp.81-96. Un problème lié à l'exploitation de graphe de connaissances, en particulier lors de traitements avec des méthodes d'apprentissage automatique, est le passage à l'échelle. Nous proposons ici une méthode pour réduire significativement la taille des graphes utilisés pour se focaliser sur une partie utile dans un contexte d'usage donné. Nous définissons ainsi la notion de graphe de contexte comme un extrait d'une ou plusieurs bases de connaissances généralistes (tels que DBpedia, Wikidata, Yago) qui contient l'ensemble d'informations pertinentes pour un domaine spécifique tout en préservant les propriétés du graphe d'origine. Nous validons l'approche sur un extrait de DBpedia pour des entités en lien avec le projet Data&Musée et le jeu de référence KORE selon deux aspects : la couverture du graphe de contexte et la préservation de la similarité entre ses entités.
• Multiplexing Gain Region of Sectorized Cellular Networks with Mixed Delay Constraints
  
  • Nikbakht Homa
  • Wigger Michèle
  • Shamai Shlomo
  , 2019, pp.1-5. The sectorized hexagonal model with mixed delay constraints is considered when both neighbouring mobile users and base stations can cooperate over rate-limited links. Each message is a combination of independent "fast" and "slow" bits, where the former are subject to a stringent delay constraint and cannot profit from cooperation. Inner and outer bounds on the multiplexing gain region are derived. The obtained results show that for small cooperation prelogs or moderate "fast" multiplexing gains, the overall performance (sum multiplexing gain) is hardly decreased by the stringent delay constraints on the "fast" bits. For large cooperation prelogs and large "fast" multiplexing gains, increasing the "fast" multiplexing gain by ∆ comes at the expense of decreasing the sum multiplexing gain by 3∆. (10.1109/SPAWC.2019.8815417)
  DOI : 10.1109/SPAWC.2019.8815417
• Managing Agent's Impression Based on User's Engagement Detection
  
  • Mancini Maurizio
  • Biancardi Beatrice
  • Dermouche Soumia
  • Lerner Paul
  • Pelachaud Catherine
  , 2019, pp.209-211. (10.1145/3308532.3329442)
  DOI : 10.1145/3308532.3329442
• Integrating Argumentation with Social Conversation between Multiple Virtual Coaches
  
  • Kantharaju Reshmashree B
  • Simonetti Donatella
  • Heylen Dirk
  • Op den Akker Harm
  • Pease Alison
  • Reidsma Dennis
  • Pelachaud Catherine
  • Snaith Mark
  • Bruijnes Merijn
  • Klaassen Randy
  • Beinema Tessa
  • Huizing Gerwin
  , 2019, pp.203-205. (10.1145/3308532.3329450)
  DOI : 10.1145/3308532.3329450
• Boosting decision stumps for dynamic feature selection on data streams
  
  • Barddal Jean Paul
  • Enembreck Fabrició
  • Gomes Heitor Murilo
  • Bifet Albert
  • Pfahringer Bernhard
  Information Systems, Elsevier, 2019, 83, pp.13-29. Feature selection targets the identification of which features of a dataset are relevant to the learning task. It is also widely known and used to improve computation times, reduce computation requirements, and to decrease the impact of the curse of dimensionality and enhancing the generalization rates of classifiers. In data streams, classifiers shall benefit from all the items above, but more importantly, from the fact that the relevant subset of features may drift over time. In this paper, we propose a novel dynamic feature selection method for data streams called Adaptive Boosting for Feature Selection (ABFS). ABFS chains decision stumps and drift detectors, and as a result, identifies which features are relevant to the learning task as the stream progresses with reasonable success. In addition to our proposed algorithm, we bring feature selection-specific metrics from batch learning to streaming scenarios. Next, we evaluate ABFS according to these metrics in both synthetic and real-world scenarios. As a result, ABFS improves the classification rates of different types of learners and eventually enhances computational resources usage. (10.1016/j.is.2019.02.003)
  DOI : 10.1016/j.is.2019.02.003
• On the CVP for the root lattices via folding with deep ReLU neural networks
  
  • Corlay Vincent
  • Boutros Joseph
  • Ciblat Philippe
  • Brunel Loïc
  , 2019. Point lattices and their decoding via neural networks are considered in this paper. Lattice decoding in R n , known as the closest vector problem (CVP), becomes a classification problem in the fundamental parallelotope with a piecewise linear function defining the boundary. Theoretical results are obtained by studying root lattices. We show how the number of pieces in the boundary function reduces dramatically with folding, from exponential to linear. This translates into a two-layer ReLU network requiring a number of neurons growing exponentially in n to solve the CVP, whereas this complexity becomes polynomial in n for a deep ReLU network.
• Enhanced chaotic performance with optically injected quantum cascade lasers
  
  • Spitz O
  • Herdt A
  • Carras M.
  • Elsässer W
  • Grillot F.
  , 2019. Quantum cascade lasers, which are able to emit deterministic chaotic patterns, are found to exhibit improved chaos properties when using optical injection instead of conventional optical feedback. These findings are of paramount importance for developing high-speed secure communications in the mid-infrared domain. (10.1109/PHOSST.2019.8794900)
  DOI : 10.1109/PHOSST.2019.8794900
• Soft-Error Vulnerability Estimation Approach Based on the SET Susceptibility of Each Gate
  
  • Batagin Armelin Fábio
  • Naviner Lirida
  • D’amore Roberto
  Electronics, MDPI, 2019, 8 (7), pp.749. (10.3390/electronics8070749)
  DOI : 10.3390/electronics8070749
• METHODS AND DEVICES FOR MULTI-CORE FIBER DATA TRANSMISSION USING DATA PRECODING
  
  • Rekaya-Ben Othman Ghaya
  • Abouseif Akram
  , 2019, pp.29. 
• DEVICES AND METHODS FOR MACHINE LEARNING ASSISTED SPHERE DECODING
  
  • Rekaya-Ben Othman Ghaya
  • Askri Aymen
  , 2019, pp.41. A decoder for decoding a signal received through a transmission channel represented by a channel matrix using a search sphere radius. The decoder comprises a radius determination device (307) for determining a search sphere radius from a preliminary radius. The radius determination device (307) is configured to: i. apply a machine learning algorithm to input data derived from the received signal, the 10 channel matrix and a current radius, the current radius being initially set to the preliminary radius, which provides a current predicted number of lattice points associated with the current radius; ii. compare the current predicted number of lattice points to a given threshold; iii. update the current radius if the current predicted number of lattice points is strictly higher 15 than the given threshold, the current radius being updated by applying a linear function to the current radius; Steps i to iii are iterated until a termination condition is satisfied, the termination condition being related to the current predicted number, the radius determination device (307) being configured to set the search sphere radius to the current radius in response to the 20 termination condition being satisfied.