Publications

2020

• Online Payments by Merely Broadcasting Messages
  
  • Collins Daniel
  • Guerraoui Rachid
  • Komatovic Jovan
  • Kuznetsov Petr
  • Monti Matteo
  • Pavlovic Matej
  • Pignolet Yvonne Anne
  • Seredinschi Dragos-Adrian
  • Tonkikh Andrei
  • Xygkis Athanasios
  , 2020.
• La longue marche des épiceries coopératives
  
  • Ouahab Alban
  , 2020. La longue marche des épiceries coopératives Alban Ouahab 12-15 minutes Le développement ces dernières années en France d'épiceries et de supermarchés coopératifs et participatifs est une nouvelle étape dans une histoire bien particulière. Leur succès actuel est le résultat d'une évolution complexe, hésitante, faite d'accélérations et de ralentissements. La forme des coopératives s'est transformée de même que les aspirations des coopérateurs et coopératrices qui sont passées de préoccupation strictement économique à la prise en compte d'enjeux sociaux et environnementaux plus vastes. Ce détour historique nous permet de comprendre que les alternatives ne s'établissent pas sans conflit, sans tumulte et sans remise en cause de leur propre modèle.
• Conférence Nationale d'Intelligence Artificielle Année 2020
  
  • Bloch Isabelle
  • Bouraoui Zied
  • Brunessaux Stephan
  • Doutre Sylvie
  • El Fallah-Seghrouchni Amal
  • Ferré Sébastien
  • Maris Frédéric
  • Niveau Alexandre
  • Sabouret Nicolas
  • Demazeau Yves
  • Longin Dominique
  , 2020.
• Méthode d’analyse sémantique d’images combinant apprentissage profond et relations structurelles par appariement de graphes
  
  • Chopin Jérémy
  • Fasquel Jean-Baptiste
  • Mouchère Harold
  • Bloch Isabelle
  • Dahyot Rozenn
  , 2020.
• Deep kernel representation learning for complex data and reliability issues
  
  • Laforgue Pierre
  , 2020. The first part of this thesis aims at exploring deep kernel architectures for complex data. One of the known keys to the success of deep learning algorithms is the ability of neural networks to extract meaningful internal representations. However, the theoretical understanding of why these compositional architectures are so successful remains limited, and deep approaches are almost restricted to vectorial data. On the other hand, kernel methods provide with functional spaces whose geometry are well studied and understood. Their complexity can be easily controlled, by the choice of kernel or penalization. In addition, vector-valued kernel methods can be used to predict kernelized data. It then allows to make predictions in complex structured spaces, as soon as a kernel can be defined on it.The deep kernel architecture we propose consists in replacing the basic neural mappings functions from vector-valued Reproducing Kernel Hilbert Spaces (vv-RKHSs). Although very different at first glance, the two functional spaces are actually very similar, and differ only by the order in which linear/nonlinear functions are applied. Apart from gaining understanding and theoretical control on layers, considering kernel mappings allows for dealing with structured data, both in input and output, broadening the applicability scope of networks. We finally expose works that ensure a finite dimensional parametrization of the model, opening the door to efficient optimization procedures for a wide range of losses.The second part of this thesis investigates alternatives to the sample mean as substitutes to the expectation in the Empirical Risk Minimization (ERM) paradigm. Indeed, ERM implicitly assumes that the empirical mean is a good estimate of the expectation. However, in many practical use cases (e.g. heavy-tailed distribution, presence of outliers, biased training data), this is not the case.The Median-of-Means (MoM) is a robust mean estimator constructed as follows: the original dataset is split into disjoint blocks, empirical means on each block are computed, and the median of these means is finally returned. We propose two extensions of MoM, both to randomized blocks and/or U-statistics, with provable guarantees. By construction, MoM-like estimators exhibit interesting robustness properties. This is further exploited by the design of robust learning strategies. The (randomized) MoM minimizers are shown to be robust to outliers, while MoM tournament procedure are extended to the pairwise setting.We close this thesis by proposing an ERM procedure tailored to the sample bias issue. If training data comes from several biased samples, computing blindly the empirical mean yields a biased estimate of the risk. Alternatively, from the knowledge of the biasing functions, it is possible to reweight observations so as to build an unbiased estimate of the test distribution. We have then derived non-asymptotic guarantees for the minimizers of the debiased risk estimate thus created. The soundness of the approach is also empirically endorsed.
• Emerging DSP techniques for multi-core fiber transmission systems
  
  • Abouseif Akram
  , 2020. Optical communication systems have seen several phases in the last decades. It is predictable that the optical systems as we know will reach the non-linear capacity limits. At the moment, the space is the last degree of freedom to be implemented in order to keep delivering the upcoming capacity demands for the next years. Therefore, intensive researches are conducted to explore all the aspects concerning the deployment of the space-division multiplexing (SDM) system. Several impairments impact the SDM systems as a result from the interaction of the spatial channels which degrades the system performance. In this thesis, we focus on the multi-core fibers (MCFs) as the most promising approach to be the first representative of the SDM system. We present different digital and optical solutions to mitigate the non-unitary effect known as the core dependent loss (CDL). The first part is dedicated to study the performance of the MCF transmission taking into account the propagating impairments that impact the MCF systems. We propose a channel model that helps to identify the MCFs system. The second part is devoted to optical technique to enhance the transmission performance with an optimal solution. After, we introduced digital techniques for further enhancement, the Zero Forcing pre-compensation and the space-time coding for further CDL mitigation. All the simulation results are validated analytically by deriving the error probability upper bounds.
• Détection de structures linéiques dans les images radar à synthèse d'ouverture par test de ratio de vraisemblance généralisé
  
  • Nicolas Gasnier
  • Denis Loïc
  • Tupin Florence
  , 2020.
• Study of hybrid silicon quantum dot frequency comb laser dynamic for 5G and datacom applications
  
  • Dong Bozhang
  • Duan Jianan
  • Huang Heming
  • Kurczveil Geza
  • Liang Di
  • Grillot Frédéric
  , 2020. This work reports on the high performance of a 1.3 µm hybrid quantum dot frequency comb laser. The material parameters such as gain, differential gain, and linewidth enhancement factor are studied and linked to the comb dynamics. In particular, results show that a larger linewidth enhancement factor is more beneficial for comb operation; moreover, we demonstrate that, by employing optical injection, both the 3-dB bandwidth and the flatness of the whole optical frequency comb is improved. Such novel findings give promising guidelines for the development of high-speed dense wavelength division multiplexing photonic integrated circuits in upcoming 5G telecommunications and datacom applications.
• When does Partial Noisy Feedback Enlarge the Capacity of a Gaussian Broadcast Channel?
  
  • Ravi Aditya Narayan
  • Pillai Sibi Raj B.
  • Prabhakaran Vinod M
  • Wigger Michèle
  , 2020, pp.1480-1485. Feedback is known to enlarge the capacity region of a Gaussian Broadcast Channel (GBC) with independent noise realizations at the receivers, and an average power constraint at the transmitter. The capacity enlargement may occur even when there is noisy feedback from only one of the two receivers. However, recent results show the existence of a feedback noise threshold, beyond which one-sided feedback from only the stronger receiver is futile in enlarging the capacity region. The current paper presents a tight characterization of the feedback noise threshold, which separates the regimes where feedback from only the stronger receiver enlarges the capacity or leaves it unchanged. The scheme used to prove this result also leads to some interesting observations on noisy feedback from only the weak receiver. (10.1109/ISIT44484.2020.9174173)
  DOI : 10.1109/ISIT44484.2020.9174173
• O (log log n) Worst-Case Local Decoding and Update Efficiency for Data Compression
  
  • Vatedka Shashank
  • Chandar Venkat
  • Tchamkerten Aslan
  , 2020, pp.2371-2376. (10.1109/ISIT44484.2020.9173968)
  DOI : 10.1109/ISIT44484.2020.9173968
• Some Results on the Vector Gaussian Hypothesis Testing Problem
  
  • Escamilla Pierre
  • Zaidi Abdellatif
  • Wigger Michèle
  , 2020. This paper studies the problem of discriminating two multivariate Gaussian distributions in a distributed manner. Specifically, it characterizes in a special case the optimal type-II error exponent as a function of the available communication rate. As a side-result, the paper also presents the optimal type-II error exponent of a slight generalization of the hypothesis testing against conditional independence problem where the marginal distributions under the two hypotheses can be different. (10.1109/ISIT44484.2020.9173998)
  DOI : 10.1109/ISIT44484.2020.9173998
• Tetrahedral Coding and Non-Unitary Resilience in Polarization-Multiplexed Lightwave Transmissions
  
  • Dumenil Arnaud
  • Awwad Elie
  • Measson Cyril
  , 2020, pp.309-314. (10.1109/ISIT44484.2020.9174068)
  DOI : 10.1109/ISIT44484.2020.9174068
• Implications of the Multi-Modality of User Perceived Page Load Time
  
  • Salutari Flavia
  • Hora Diego Da
  • Varvello Matteo
  • Teixeira Renata
  • Christophides Vassilis
  • Rossi D.
  , 2020, pp.1-8. Web browsing is one of the most popular applications for both desktop and mobile users. A lot of effort has been devoted to speedup the Web, as well as in designing metrics that can accurately tell whether a webpage loaded fast or not. An often implicit assumption made by industrial and academic research communities is that a single metric is sufficient to assess whether a webpage loaded fast. In this paper we collect and make publicly available a unique dataset which contains webpage features (e.g., number and type of embedded objects) along with both objective and subjective Web quality metrics. This dataset was collected by crawling over 100 websites-representative of the top 1 M websites in the Web-while crowdsourcing 6,000 user opinions on user perceived page load time (uPLT). We show that the uPLT distribution is often multi-modal and that, in practice, no more than three modes are present. The main conclusion drawn from our analysis is that, for complex webpages, each of the different objective QoE metrics proposed in the literature (such as AFT, TTI, PLT, etc.) is suited to approximate one of the different uPLT modes. (10.1109/MedComNet49392.2020.9191615)
  DOI : 10.1109/MedComNet49392.2020.9191615
• Digital Interpolating Phase Modulator Implementation for Outphasing PA
  
  • Franco Gabriel Souza
  • Pham Dang-Kièn Germain
  • Desgreys Patricia
  , 2020, pp.259-262. (10.1109/NEWCAS49341.2020.9159834)
  DOI : 10.1109/NEWCAS49341.2020.9159834
• Online Depth Learning Against Forgetting in Monocular Videos
  
  • Zhang Zhenyu
  • Lathuilière Stéphane
  • Ricci Elisa
  • Sebe Nicu
  • Yan Yan
  • Yang Jian
  , 2020. Online depth learning is the problem of consistently adapting a depth estimation model to handle a continuously changing environment. This problem is challenging due to the network easily overfits on the current environment and forgets its past experiences. To address such problem , this paper presents a novel Learning to Prevent Forgetting (LPF) method for online mono-depth adaptation to new target domains in unsupervised manner. Instead of updating the universal parameters, LPF learns adapter modules to efficiently adjust the feature representation and distribution without losing the pre-learned knowledge in online condition. Specifically, to adapt temporal-continuous depth patterns in videos, we introduce a novel meta-learning approach to learn adapter modules by combining online adaptation process into the learning objective. To further avoid overfitting, we propose a novel temporal-consistent regu-larization to harmonize the gradient descent procedure at each online learning step. Extensive evaluations on real-world datasets demonstrate that the proposed method, with very limited parameters, significantly improves the estimation quality.
• Feature selection algorithms for flexible analog-to-feature converter
  
  • Back Antoine
  • Chollet Paul
  • Fercoq Olivier
  • Desgreys Patricia
  , 2020, pp.186-189. One of the main challenges in the field of wireless sensors is to increase their battery life. Analog-to-feature (A2F) conversion is an acquisition method thought for IoT devices, that perform classification tasks at sub-Nyquist rate, by extracting relevant features in the analog domain and then performing the classification step in the digital domain. Current A2F solutions are designed for a specific application, this paper proposes a method to design a generic A2F converter usable for several signal types. In order to extract information for classification task, we propose to use non uniform wavelet sampling, its drawback is that it brings redundancy and irrelevant information. To reach our goal of decreasing power consumption, we need to extract a small set of relevant features for classification. To achieve this, several features selection algorithms are tested for electrocardiogram (ECG) anomalies detection. We demonstrate that the detection rate of ECG anomalies can reach 98% with less than 10 features extracted. (10.1109/NEWCAS49341.2020.9159817)
  DOI : 10.1109/NEWCAS49341.2020.9159817
• Detection-Localization-Identification of Vibrations Over Long Distance SSMF With Coherent $\Delta \phi$-OTDR
  
  • Awwad Elie
  • Dorize Christian
  • Guerrier Sterenn
  • Renaudier Jeremie
  Journal of Lightwave Technology, Institute of Electrical and Electronics Engineers (IEEE)/Optical Society of America(OSA), 2020, 38 (12), pp.3089-3095. (10.1109/JLT.2020.2993167)
  DOI : 10.1109/JLT.2020.2993167
• Stochastic D2D Caching with Energy Harvesting Nodes
  
  • Nikbakht Homa
  • Kamel Sarah
  • Wigger Michèle
  • Yener Aylin
  , 2020. Consider a stochastic wireless device-to-device (D2D) caching network with nodes that are harvesting energy from external sources at random times. Each node is equipped with a cache memory, where the node prefetches maximum distance separable (MDS) coded packets of the files from a given library. When a node requests a file from this library, neighbouring nodes are asked to send the relevant missing subfiles over noisy channels. This work presents different selection strategies to determine which neighbouring nodes should transmit which missing subfiles. The strategies can roughly be divided into three categories: sequential strategies where transmission stops when the requesting node has correctly decoded enough subfiles; coordinated strategies where the requesting node is informed about the other nodes' cache contents and centrally decides which node should send which file; and adaptive strategies where the requesting node sequentially decides on which files should be sent in function of the subfiles that it previously decoded correctly. Our numerical simulations show that at moderate energy levels or when there are many file requests, sequential strategies perform significantly worse than coordinated or adaptive strategies. On the other hand, at high energy levels sequential strategies (or even completely decentralized strategies) perform as well or even better. These latter strategies should thus be prefered as they come with less synchronization overhead and delay. The same applies for environments with only few transmission errors (i.e., in high quality channels).
• Distributed Hypothesis Testing with Variable-Length Coding
  
  • Salehkalaibar Sadaf
  • Wigger Michèle
  , 2020. This paper characterizes the optimal type-II error exponent for a distributed hypothesis testing-against-independence problem when the expected rate of the sensor-detector link is constrained. Unlike for the well-known Ahlswede-Csiszar result that holds under a maximum rate constraint and where a strong converse holds, here the optimal exponent depends on the allowed type-I error exponent. Specifically, if the type-I error probability is limited by , then the optimal type-II error exponent under an expected rate constraint R coincides with the optimal type-II error exponent under a maximum rate constraint of (1 −)R.
• On the Optimization of Recursive Relational Queries: Application to Graph Queries
  
  • Jachiet Louis
  • Genevès Pierre
  • Gesbert Nils
  • Layaïda Nabil
  , 2020, pp.1-23. Graph databases have received a lot of attention as they are particularly useful in many applications such as social networks, life sciences and the semantic web. Various languages have emerged to query graph databases, many of which embed forms of recursion which reveal essential for navigating in graphs. The relational model has benefited from a huge body of research in the last half century and that is why many graph databases rely on techniques of relational query engines. Since its introduction, the relational model has seen various attempts to extend it with recursion and it is now possible to use recursion in several SQL or Datalog based database systems. The optimization of recursive queries remains, however, a challenge. We propose mu-RA, a variation of the Relational Algebra equipped with a fixpoint operator for expressing recursive relational queries. mu-RA can notably express unions of conjunctive regular path queries. Leveraging the fact that this fixpoint operator makes recursive terms more amenable to algebraic transformations, we propose new rewrite rules. These rules makes it possible to generate new query execution plans, that cannot be obtained with previous approaches. We present the syntax and semantics of mu-RA, and the rewriting rules that we specifically devised to tackle the optimization of recursive queries. We report on practical experiments that show that the newly generated plans can provide significant performance improvements for evaluating recursive queries over graphs. (10.1145/3318464.3380567)
  DOI : 10.1145/3318464.3380567
• Creating DALI, a Large Dataset of Synchronized Audio, Lyrics, and Notes
  
  • Meseguer-Brocal Gabriel
  • Cohen-Hadria Alice
  • Peeters Geoffroy
  Transactions of the International Society for Music Information Retrieval (TISMIR), Ubiquity Press, 2020, 3 (1), pp.55-67. (10.5334/tismir.30)
  DOI : 10.5334/tismir.30
• Improvement of the angle of arrival measurement accuracy for indoor UWB localization
  
  • Awarkeh Nour
  • Cousin Jean-Christophe
  • Muller Muriel
  • Samama Nel
  Journal of Sensors, Hindawi Publishing Corporation, 2020, 2020, pp.2603861:1-2603861:8. This paper shows that the accuracy of azimuth angle measurement for an interferometric localization system used to locate tags in its Line-of-Sight (LoS) can be improved by exploiting Impulse Radio-Ultra WideBand (IR-UWB) signals and without increasing the frequency bandwidth. This solution uses a Phase Correlation (PC) method, initially applied for Continuous Wave (CW) signals, adapted for Ultra WideBand (UWB) pulse signals. The obtained results are compared to those computed by a classical Energy Detection (ED) method where it becomes impossible to estimate azimuth angles for tag positions close to the orthogonal centered axis of the localization system baseline. (10.1155/2020/2603861)
  DOI : 10.1155/2020/2603861
• Precise and Efficient Analysis of Context-Sensitive Cache Conflict Sets
  
  • Brandner Florian
  • Noûs Camille
  , 2020, pp.44-55. (10.1145/3394810.3394811)
  DOI : 10.1145/3394810.3394811
• Enabling adaptive bitrate algorithms in hybrid CDN/P2P networks
  
  • Yousef Hiba
  • Le Feuvre Jean
  • Ageneau Paul-Louis
  • Storelli Alexandre
  , 2020, pp.54-65. As video traffic becomes the dominant part of the global Internet traffic, keeping a good quality of experience (QoE) becomes more challenging. To improve QoE, HTTP adaptive streaming with various adaptive bitrate (ABR) algorithms has been massively deployed for video delivery. Based on their required input information, these algorithms can be classified, into buffer-based, throughput-based or hybrid buffer-throughput algorithms. Nowadays, due to their low cost and high scalability, peer-to-peer (P2P) networks have become an efficient alternative for video delivery over the Internet, and many attempts at merging HTTP adaptive streaming and P2P networks have surfaced. However, the impact of merging these two approaches is still not clear enough, and interestingly, the existing HTTP adaptive streaming algorithms lack testing in a P2P environment. In this paper, we address and analyze the main problems raised by the use of the existing HTTP adaptive streaming algorithms in the context of P2P networks. We propose two method-ologies to make these algorithms more efficient in P2P networks regardless of the ABR algorithm used, one favoring overall QoE and one favoring P2P efficiency. Additionally, we propose two new metrics to quantify the P2P efficiency for ABR delivery over P2P. CCS CONCEPTS • Information systems → Multimedia streaming. (10.1145/3339825.3391859)
  DOI : 10.1145/3339825.3391859
• Optimal probing sequences for polarization-multiplexed coherent phase OTDR
  
  • Dorize Christian
  • Awwad Elie
  • Guerrier Sterenn
  • Renaudier Jérémie
  , 2020, pp.T3.23. (10.1364/OFS.2020.T3.23)
  DOI : 10.1364/OFS.2020.T3.23