Publications

2018

• Infinite-Task Learning with Vector-Valued RKHSs
  
  • Brault Romain
  • Lambert Alex
  • Szabo Zoltan
  • Sangnier Maxime
  • d'Alché-Buc Florence
  , 2018. Machine learning has witnessed the tremendous success of solving tasks depending on a hyperparameter. While multi-task learning is celebrated for its capacity to solve jointly a finite number of tasks, learning a continuum of tasks for various loss functions is still a challenge. A promising approach, called Parametric Task Learning, has paved the way in the case of piecewise-linear loss functions. We propose a generic approach, called Infinite-Task Learning, to solve jointly a continuum of tasks via vector-valued RKHSs. We provide generalization guarantees to the suggested scheme and illustrate its efficiency in cost-sensitive classification, quantile regression and density level set estimation.
• Shaping photon-pair time-frequency correlations in inhibited-coupling hollow-core fibers
  
  • Cordier Martin
  • Orieux Adeline
  • Debord Benoît
  • Gérôme Frédéric
  • Gorse A.
  • Chafer Matthieu
  • Diamanti Eleni
  • Delaye Philippe
  • Benabid Fetah
  • Zaquine Isabelle
  , 2018.
• Parallel-beam ROI reconstruction with differentiated backprojection and angularly subsampled complementary sinograms
  
  • Reshef Aymeric
  • Nikoukhah Tina
  • Riddell Cyril
  • Trousset Yves
  • Ladjal Saïd
  • Bloch Isabelle
  , 2018. Recently, we introduced a parallel-beam two-pass analytical reconstruction that allows truncation to be accounted for in the image domain rather than the projection domain. In particular, we showed that backprojection of a vastly angularly undersampled sinogram of un-truncated data could be used to extrapolate the backprojection of a finely sampled, fully truncated sinogram of the same object to perform more accurate region-of-interest (ROI) imaging. The same extrapolation idea can be performed using differentiated backprojection (DBP). The goal of this study is to give a general DBP-based formula when reconstructing a finite set of projections in parallel geometry. We discuss the discretization of this formula, in particular when the image grid size is large with respect to the number of projections, and we show how it can be applied to our extrapolation problem.
• A Stochastic Model for Vehicle Clustering Performance Analysis
  
  • Ren Mengying
  • Zhang Jun
  • Khoukhi Lyes
  • Labiod Houda
  • Vèque Véronique
  , 2018, pp.1-6. (10.1109/icc.2018.8422606)
  DOI : 10.1109/icc.2018.8422606
• How to Split UL/DL Antennas in Full-Duplex Cellular Networks
  
  • da Silva Jose Mairton B.
  • Ghauch Hadi
  • Fodor Gabor
  • Fischione Carlo
  , 2018, pp.1-6. (10.1109/ICCW.2018.8403645)
  DOI : 10.1109/ICCW.2018.8403645
• Optimal Resource Scheduling for Energy Harvesting Communications under Strict Delay Constraint
  
  • Fawaz Ibrahim
  • Sarkiss Mireille
  • Ciblat Philippe
  , 2018. This paper investigates the resource scheduling minimizing the packet loss when the wireless communication system operates with Energy Harvesting (EH) based devices. The packet loss occurs when the buffer is overflowed and when the queued packet is older than a certain pre-defined threshold. We so consider a strict delay constraint rather than an average delay constraint. The associated optimization problem can be modeled as Markov Decision Problem (MDP) where the actions are the number of packets sent on the known channel at each slot. The optimal deterministic offline policy is exhibited through dynamic programming techniques, i.e. Value Iteration (VI) algorithm. We show the gain in the number of transmitted packets and the consumed energy is substantial compared to a naive policy which forces the system to send the maximum number of packets using the available energy in the battery. (10.1109/ICC.2018.8422088)
  DOI : 10.1109/ICC.2018.8422088
• Optimal Joint Subcarrier and Power Allocation in NOMA is Strongly NP-Hard
  
  • Salaun Lou
  • Chen Chung Shue
  • Coupechoux Marceau
  , 2018, pp.1-7. Non-orthogonal multiple access (NOMA) is a promising radio access technology for 5G. It allows several users to transmit on the same frequency and time resource by performing power-domain multiplexing. At the receiver side, successive interference cancellation (SIC) is applied to mitigate interference among the multiplexed signals. In this way, NOMA can outperform orthogonal multiple access schemes used in conventional cellular networks in terms of spectral efficiency and allows more simultaneous users. This paper investigates the computational complexity of joint subcarrier and power allocation problems in multi-carrier NOMA systems. We prove that these problems are strongly NP-hard for a large class of objective functions, namely the weighted generalized means of the individual data rates. This class covers the popular weighted sum-rate, proportional fairness, harmonic mean and max-min fairness utilities. Our results show that the optimal power and subcarrier allocation cannot be computed in polynomial time in the general case, unless P = NP. Nevertheless, we present some tractable special cases and we show that they can be solved efficiently. (10.1109/ICC.2018.8422362)
  DOI : 10.1109/ICC.2018.8422362
• Low-Overhead Coordination in Sub-28 Millimeter-Wave Networks
  
  • Ghauch Hadi
  • Kim Taejoon
  • Skoglund Mikael
  • Fischione Carlo
  , 2018, pp.1-6. (10.1109/ICC.2018.8422456)
  DOI : 10.1109/ICC.2018.8422456
• Beyond Bits: A Quaternary FPGA Architecture Using Multi-Vt Multi-Vdd FDSOI Devices
  
  • Chaudhuri Sumanta
  , 2018, pp.38-43. In this article we present the architecture of a quaternary FPGA, its implementation in FDSOI technology, and a comparison with binary architectures based on VPR. We discuss the transistor level design of LUTs, Flip-Flops, Muxes, and multi-valued buffer circuits exploiting the capability of FDSOI technology to modify threshold voltages. We present I/O elements of such an FPGA with binary to quaternary translators and a new technique to reduce global routing by combining clock and reset on the same wire. We model the area, delay and power consumption of the quaternary FPGA architecture in VPR. We compare the implementation of very simple arithmetic benchmarks with equivalent two-valued FPGA architectures in VPR. We show that, it is possible to achieve upto 15% reduction in transistor area, and 10% reduction in critical path delay. (10.1109/ISMVL.2018.00015)
  DOI : 10.1109/ISMVL.2018.00015
• Stochastic models for cellular networks planning and performance assessment
  
  • Gomez Jean-Sébastien
  , 2018. With the booming of the ubiquitous and nomad Internet of Things, wireless systems and networks must support the limitless development of a digitalized eco-system. Being the backbone of the connected devices, asserting and optimizing wireless network performance is of a major importance. This dissertation aims at introducing methods and numerical analysis frameworks to enhance our comprehension of performance at a network level. Indeed, network performance has been widely explored from the point of view of the link capacity. Thanks to stochastic geometry and point process, we are able characterize the influence of the positions of the antennas. In this dissertation, the \beta-Ginibre point process is chosen to model the locations of the base stations in the plain. The \beta-Ginibre is a repulsive point process in which repulsion is controlled by the parameter. When \beta tends to zero, the point process converges in law towards a Poisson point process. If \beta equals to one it becomes a Ginibre point process. Simulations on real data collected in France show that base station locations can be fitted with a \beta-Ginibre point process. Moreover we prove that their superposition tends to a Poisson point process as it can be seen from real data. Qualitative interpretations on deployment strategies are derived from the model fitting of the raw data. The parameter that represents the deployment strategy of a operator, is also an indicator of the overall signal quality in the network : the more regular the deployment is, the better the overall signal quality. In order to quantify the gain in performance induced by a higher , a interference limited network model based on marked point process and loss probability has been introduced. In order to generalize performance analysis to any networks, a scheme based on the Cournot-Nash equilibria is investigated. Under this general framework, only the signal quality between nodes is required to derive a resource allocation strategy for the overall network. Supply in resources of the network and traffic requirements are modeled by probability measures. The optimal resource allocation strategy is derived by the coupling between the two probability measures that minimizes a specific quadratic objective function. Numerical analysis highlights that there exists an optimal working point, where users satisfaction and network occupancy are equal.
• On Achievability for Downlink Cloud Radio Access Networks with Base Station Cooperation
  
  • Wang Chien-Yi
  • Wigger Michèle
  • Zaidi Abdellatif
  IEEE Transactions on Information Theory, Institute of Electrical and Electronics Engineers, 2018, 64 (8), pp.5726-5742. This work investigates the downlink of a cloud radio access network (C-RAN) in which a central processor communicates with two mobile users through two base stations (BSs). The BSs act as relay nodes and cooperate with each other through error-free rate-limited links. We develop and analyze two coding schemes for this scenario. The first coding scheme is based on Liu-Kang scheme for C-RANs without BS cooperation; and extends it to scenarios allowing conferencing between the BSs. Among few other features, our new coding scheme enables arbitrary correlation among the auxiliary codewords that are recovered by the BSs. It also introduces common codewords to be described to both BSs. For the analysis of this coding scheme, we extend the multivariate covering lemma to non-Cartesian product sets, thereby correcting an erroneous application of this lemma in Liu-Kang's related work. We highlight key aspects of this scheme by studying three important instances of it. The second coding scheme extends the so-called compression scheme that was originally developed for memoryless Gaussian C-RANs without BS cooperation to general discrete memoryless C-RANs with BS cooperation. We show that this scheme subsumes the original compression scheme when applied to memoryless Gaussian C-RAN models. In the analysis of this scheme, we also highlight important connections with the so-called distributed decode--forward scheme, and refine the approximate capacity of a general $N$-BS $L$-user C-RAN model in the memoryless Gaussian case. (10.1109/TIT.2018.2836675)
  DOI : 10.1109/TIT.2018.2836675
• An M : N Shared Regenerator Protection Scheme in Translucent WDM Networks
  
  • Doumith Elias
  • Al Zahr Sawsan
  , 2018, pp.28-36. (10.23919/IFIPNetworking.2018.8696420)
  DOI : 10.23919/IFIPNetworking.2018.8696420
• Shaping photon-pairs time-frequency correlations in inhibited-coupling hollow-core fibers
  
  • Cordier Martin
  • Orieux Adeline
  • Debord Benoît
  • Gérôme Frédéric
  • Gorse A.
  • Chafer Matthieu
  • Diamanti Eleni
  • Delaye Philippe
  • Benabid Fetah
  • Zaquine Isabelle
  , 2018, pp.1-2. We experimentally show how multiband dispersion properties of inhibited-coupling hollow-core fibers allow to control the spectral correlations of photon pairs generated through four-wave-mixing in a fiber filled with non-linear gas. (10.1364/CLEO_QELS.2018.FM4G.4)
  DOI : 10.1364/CLEO_QELS.2018.FM4G.4
• Mode transformation in an inhibited-coupling guiding asymmetric tubular hollow fiber
  
  • Osorio Jonas
  • Chafer Matthieu
  • Debord Benoît
  • Giovanardi Fabio
  • Cordier Martin
  • Delahaye Frédéric, Lucien
  • Vincetti Luca
  • Gérôme Frédéric
  • Benabid Fetah
  , 2018, pp.SF1K.6.
• On Helping and Stacks
  
  • Aksenov Vitalii
  • Kuznetsov Petr
  • Shalyto Anatoly
  , 2018. A concurrent algorithm exhibits helping when one process performs work on behalf of other processes. More formally, helping is observed when the order of some operation in a linearization is fixed by a step of another process. In this paper, we show that no wait-free linearizable implementation of a stack using read, write, compare&swap and fetch&add operations can be help-free, correcting a mistake in an earlier proof by Censor-Hillel et al.
• Cooperative relay beamforming for control channel jamming in vehicular networks
  
  • Gu Pengwenlong
  • Hua Cunqing
  • Khatoun Rida
  • Wu Yue
  • Serhrouchni Ahmed
  , 2018, pp.1-7. (10.23919/WIOPT.2018.8362860)
  DOI : 10.23919/WIOPT.2018.8362860
• Étude de techniques de modulation pour les réseaux optiques d'accès de nouvelle génération
  
  • Chaibi Mohamed Essghair
  , 2018.
• Vertebral corners detection on sagittal X-rays based on shape modelling, random forest classifiers and dedicated visual features
  
  • Ebrahimi Shahin
  • Gajny Laurent
  • Skalli Wafa
  • Angelini Elsa
  Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization, Taylor & Francis, 2018, 7 (2), pp.132-144. Quantitative measurements of spine shape parameters on planar X-ray images is critical for clinical applications but remains tedious and with no fully-automated solution demonstrated on the whole spine. This study aims to limit manual input, while demonstrating precise vertebrae corners positioning and shape parameter measurements from sagittal radiographs of the cervical and lumbar regions, exploiting novel dedicated visual features and specialized classifiers. A database of manually annotated X-ray images is used to train specialized Random Forest classifiers for each spine regions and corner types. An original combination of local gradient characteristics, Haar-like features, and contextual features based on patch intensity and contrast is used as visual features. The proposed method is evaluated on 49 sagittal X-rays of asymptomatic and pathological subjects, from multiple imaging sites, and with a large age range (6 – 69 years old). Performance is first evaluated for positioning a 2D spine shape model, where precisely detected corners enable to adjust the model via an original multilinear statistical regression. Root-mean square errors (RMSE) of corners localization and vertebra orientations are reported, demonstrating state-of-the-art precision compared to existing methods, but with minimal manual input. The method is then evaluated for the extraction of additional vertebrae shape characteristics, such as centre positioning, endplate centres positioning and endplate length measures, rarely studied in previous literature. The proposed method enables, with minimal initialization, fast and precise individual vertebrae delineations on sagittal radiographs on normal and pathological cases, with a level of precision and robustness required for objective support for diagnosis and therapy decision making. (10.1080/21681163.2018.1463174)
  DOI : 10.1080/21681163.2018.1463174
• Anycast de nos jours
  
  • Cicalese Danilo
  , 2018. Les motivations des recherches réalisées dans cette thèse viennent de la curiosité de découvrir IP anycast. Cette technique est couramment utilisée pour partager la quantité d’information d'une variété de services globaux. Dans cette thèse, nous visons à sensibiliser la communauté de l'utilisation d'IP anycast. Tout d'abord, nous nous concentrons sur l'identification d'une technique, qui ne dépend pas d'un protocole spécifique, utilisé pour la découverte et la géolocalisation des répliques anycast. D'autres techniques existent déjà, mais ils exploitent les spécificités du protocole DNS, ce qui limite leur applicabilité à ce service. Nous fournissons également à la communauté des logiciels open-source avec des jeu de données. Cette méthodologie nous a permis de mettre en oeuvre la prochaine étape de notre recherche: dévoiler toutes les entreprises qui utilisent actuellement anycast pour leurs services. Nous effectuons plusieurs IPv4 census d'anycast, en utilisant des mesures de latence à partir d'une plateforme distribuée. Ces census révèlent finalement que de nombreuses grandes entreprises d'Internet utilisent anycast. Nous avons donc décidé de mettre en place un système capable d'effectuer des census IPv4 mensuels et d'analyser les résultats. Enfin, pour compléter l'étude, nous analysons les CDN anycast. Nous effectuons une caractérisation passive en mettant l'accent sur les services qu'ils offrent et leur pénétration etc. Nos résultats révèlent que normalment plus de 50\% des internautes accèdent au contenu servi par les CDN anycast. Une large gamme de services TCP (Transmission Control Protocol) est offerte sur anycast. Celle-ci peut inclure des services d'audio, de streaming vidéo ou des HTTP & HTTPS, ces derniers étant les plus populaires. (10.70675/a29db2f7z6782z4b3dza0d3z0d7d6ae6d1d9)
  DOI : 10.70675/a29db2f7z6782z4b3dza0d3z0d7d6ae6d1d9
• Polarization-Division-Multiplexed Nonlinear Frequency Division Multiplexing
  
  • Gui Tao
  • Gemechu Wasyhun A.
  • Goossens Jen-Willem
  • Song Mengdi
  • Yousefi Mansoor
  • Hafermann Hartmut
  • Tao Lau Alan Pak
  • Jaouën Yves
  , 2018, pp.paper STu4C.3.
• Region-of-interest imaging with differentiated backprojection and angularly subsampled complementary sinograms
  
  • Reshef Aymeric
  • Nikoukhah Tina
  • Riddell Cyril
  • Tousset Yves
  • Ladjal Saïd
  • Bloch Isabelle
  , 2018. Recently, we introduced a two-pass analytical recon- struction that allows truncation to be accounted for in the image domain rather than the projection domain. In particular, we showed that backprojection of a vastly angularly undersampled sinogram of un-truncated data could be used to extrapolate the backprojection of a finely sampled, fully truncated sinogram of the same object to perform more accurate region-of-interest (ROI) imaging. The same extrapolation idea can be performed using differentiated backprojection (DBP). The goal of this study is to give a general DBP-based formula when reconstructing a finite set of projections in parallel geometry. We discuss the discretization of this formula, in particular when the image grid size is large with respect to the number of projections, and we show how it can be applied to our extrapolation problem.
• Tuple Reconstruction
  
  • Er Ngurah Agus Sanjaya
  • Ba Mouhamadou Lamine
  • Abdessalem Talel
  • Bressan Stéphane
  , 2018, pp.239-254. (10.1007/978-3-319-91455-8\_21)
  DOI : 10.1007/978-3-319-91455-8\_21
• Design Space Exploration of Magnetic Tunnel Junction based Stochastic Computing in Deep Learning
  
  • Wang You
  • Zhang Yue
  • Youguang Zhabg
  • Zhao Weisheng
  • Cai Hao
  • Naviner Lirida
  , 2018, pp.23-25.
• Coverage Analysis and Load Balancing in HetNets With Millimeter Wave Multi-RAT Small Cells
  
  • Ghatak Gourab
  • de Domenico Antonio
  • Coupechoux Marceau
  IEEE Transactions on Wireless Communications, Institute of Electrical and Electronics Engineers, 2018, 17 (5), pp.3154-3169. We characterize a two tier heterogeneous network, consisting of classical sub-6 GHz macro cells, and multi radio access technology (RAT) small cells able to operate in sub-6 GHz and millimeter-wave (mm-wave) bands. For optimizing coverage and to balance loads, we propose a two-step mechanism based on two biases for tuning the tier and RAT selection, where the sub-6 GHz band is used to speed-up the initial access procedure in the mm-wave RAT. First, we investigate the effect of the biases in terms of signal-to-interference-plus-noise ratio (SINR) distribution, cell load, and user throughput. More specifically, we obtain the optimal biases that maximize either the SINR coverage or the user downlink throughput. Then, we characterize the cell load using the mean cell approach and derive upper bounds on the overloading probabilities. Finally, for a given traffic density, we provide the small cell density required to satisfy system constraints in terms of overloading and outage probabilities. Our analysis highlights the importance of deploying dual-band small cells, in particular, when small cells are sparsely deployed or in case of heavy traffic. (10.1109/TWC.2018.2807426)
  DOI : 10.1109/TWC.2018.2807426
• TCP Congestion Control in Datacenter Optical Packet Networks on Hybrid Switches
  
  • Minakhmetov Artur
  • Ware Cédric
  • Iannone Luigi
  J. Opt. Commun. Netw., 2018, 10 (7), pp.B71-B81. Optical packet switching (OPS) has long promised performance and energy consumption improvements by doing away with optical-to-electronic conversions required by electronic packet switching; however, not having practical optical buffers makes OPS highly vulnerable to contention. This study reports on the possible and plausible use of OPS technology on datacenter networks by coupling two concepts: optical switches with shared electronic buffers, also known as hybrid switches, and the introduction of TCP congestion control algorithms (CCAs) to control transport of optical packets. The stop-and-wait (SAW) and the modified additive increase multiple decrease (mAIMD) CCAs families are reviewed. For SAW, a basic version and a modified one -- stop-and-wait-longer, adapted for hybrid switches -- are analyzed. As for mAIMD, a TCP selective acknowledgment (SACK) implementation and its simplified modification TCP mSACK are studied. It is successfully shown that these algorithms paired with the use of shared electronic buffers in hybrid switches significantly outperform bufferless all-optical switches and reach the level of all-electronic switches in datacenters and local area networks in terms of network throughput. (10.1364/JOCN.10.000B71)
  DOI : 10.1364/JOCN.10.000B71