Publications

2020

• CS-ARF: Compressed Adaptive Random Forests for Evolving Data Stream Classification
  
  • Bahri Maroua
  • Gomes Heitor Murilo
  • Bifet Albert
  • Maniu Silviu
  , 2020, pp.1-8. Ensemble-based methods are one of the most often used methods in the classification task that have been adapted to the stream setting because of their high learning performance achievement. For instance, Adaptive Random Forests (ARF) is a recent ensemble method for evolving data streams that proved to be of a good predictive performance but, as all ensemble methods, it suffers from a severe drawback related to the high computational demand which prevents it from being efficient and further exacerbates with high-dimensional data. In this context, the application of a dimensionality reduction technique is crucial while processing the Internet of Things (IoT) data stream with ultrahigh dimensionality. In this paper, we aim to alleviate this deficiency and improve ARF performance, so we introduce the CS-ARF approach that uses Compressed Sensing (CS) as an internal pre-processing task, to reduce the dimensionality of data before starting the learning process, that will potentially lead to a meaningful improvement in memory usage. Experiments on various datasets show the high classification performance of our CS-ARF approach compared against current state-of-the-art methods while reducing resource usage. (10.1109/IJCNN48605.2020.9207188)
  DOI : 10.1109/IJCNN48605.2020.9207188
• A Fully Connected Neural Network Approach to Mitigate Fiber Nonlinear Effects in 200G DP-16-QAM Transmission System
  
  • Catanese Clara
  • Ayassi Reda
  • Pincemin Erwan
  • Jaouën Yves
  , 2020, pp.1-4. (10.1109/ICTON51198.2020.9203197)
  DOI : 10.1109/ICTON51198.2020.9203197
• On Ensemble Techniques for Data Stream Regression
  
  • Gomes Heitor Murilo
  • Montiel Jacob
  • Mastelini Saulo Martiello
  • Pfahringer Bernhard
  • Bifet Albert
  , 2020, pp.1--8. An ensemble of learners tends to exceed the predictive performance of individual learners. This approach has been explored for both batch and online learning. Ensembles methods applied to data stream classification were thoroughly investigated over the years, while their regression counterparts received less attention in comparison. In this work, we discuss and analyze several techniques for generating, aggregating, and updating ensembles of regressors for evolving data streams. We investigate the impact of different strategies for inducing diversity into the ensemble by randomizing the input data (resampling, random subspaces and random patches). On top of that, we devote particular attention to techniques that adapt the ensemble model in response to concept drifts, including adaptive window approaches, fixed periodical resets and randomly determined windows. Extensive empirical experiments show that simple techniques can obtain similar predictive performance to sophisticated algorithms that rely on reactive adaptation (i.e., concept drift detection and recovery). (10.1109/IJCNN48605.2020.9206756)
  DOI : 10.1109/IJCNN48605.2020.9206756
• Performance measures for evolving predictions under delayed labelling classification
  
  • Grzenda Maciej
  • Gomes Heitor Murilo
  • Bifet Albert
  , 2020, pp.1--8. For many streaming classification tasks, the ground truth labels become available with a non-negligible latency. Given this delayed labelling setting, after the instance data arrives and before its true label is known, the online classifier model may change. Hence, the initial prediction can be replaced with additional periodic predictions gradually produced before the true label becomes available. The quality of these predictions may largely vary. Thus, the question arises of how to summarise the performance of these models when multiple predictions for a single instance are made due to delayed labels.In this study, we aim to provide intuitive performance measures summarising the performance of multiple predictions made for individual instances before their true labels arrive. Particular attention is paid to the fact that under the delayed label setting, the emphasis placed on the quality of initial predictions can vary depending on problem needs. The intermediate performance measures we propose complement existing initial and test-then-train performance evaluation when verification latency is observed. Results provided for both real and synthetic datasets show that the new measures can be used to easily rank methods in terms of their ability to produce and refine predictions before the true labels arrive. (10.1109/IJCNN48605.2020.9207256)
  DOI : 10.1109/IJCNN48605.2020.9207256
• Randomizing the Self-Adjusting Memory for Enhanced Handling of Concept Drift
  
  • Bifet Albert
  • Losing Viktor
  • Hammer Barbara
  • Wersing Heiko
  , 2020, pp.1--8. Real-time learning from data streams in non-stationary environments gains ever more relevance due to the exponentially increasing amounts of generated data. Recently, the Self-Adjusting Memory (SAM) was proposed, an algorithm able to robustly handle heterogeneous types on the basis of two dedicated memories for the current and former concepts that continuously preserve consistency with explicit filtering. Yet, since the algorithm is restricted to one memory architecture, the variety of possible alternatives is limited by design in favor of an overall model consistency. Moreover, it does not actively detect drift, thus adapting with a relatively high delay in case of abrupt changes. We propose a dynamic ensemble on the basis of the SAM algorithm, which is triggered by both, the inherent passive adaptation of SAM and active drift detection. Further, since SAM is based on the stable k-Nearest-Neighbor algorithm, we investigate multiple approaches to obtain a high diversity in the ensemble, resulting in an effective overall strategy. The increased computational demand is countered on the basis of a parallel implementation. We extensively evaluate the method on numerous benchmarks, where it consistently achieves superior results in comparison to state-of-the-art methods. (10.1109/IJCNN48605.2020.9207583)
  DOI : 10.1109/IJCNN48605.2020.9207583
• EVALUATION OF SPECIFIC ABSORPTION RATE IN THE FAR-FIELD, NEAR-TO-FAR FIELD AND NEAR-FIELD REGIONS FOR INTEGRATIVE RADIOFREQUENCY EXPOSURE ASSESSMENT
  
  • Liorni Ilaria
  • Capstick Myles
  • van Wel Luuk
  • Wiart Joe
  • Joseph Wout
  • Cardis Elisabeth
  • Guxens Mònica
  • Vermeulen Roel
  • Thielens Arno
  Radiation Protection Dosimetry, Oxford University Press (OUP), 2020, 190 (4), pp.459-472. (10.1093/rpd/ncaa127)
  DOI : 10.1093/rpd/ncaa127
• Cross-layer congestion control and quality of services in mobile networks
  
  • Zhong Zhenzhe
  , 2020. The mobile network is a hybrid network with Radio Access part and the fixed backhaul core network. The congestion control algorithms(CCA) designed for a specific type of system may not work as well in the other kind of network, especially the network with hybrid feature device like the mobile edge network. Generally, the bottleneck in a mobile network is the Radio access part. However, this is not always the case since multiple radio base stations or packet delivery network gateway can be sharing the same bottleneck in the packet delivery backhaul. In this thesis, we start from a cross-layer method and address the issue with a ubiquitous solution. The first algorithm we analysed is called CQIC, which get the PHY layer of UE involved in the congestion control design. An improvement from 3G CQIC to LTE scenario is proposed named DCIC. This algorithm uses the Downlink Control Indicator(DCI) instead of Channel Quality Indicator(CQI) to save the computation power on UE and take the scheduling result of eNB into consideration. Further, we evaluated current BBR algorithm, which focuses on the data centre network, in the mobile scenario. Most conventional CCA does not take the uplink BW degradation and other features in the cellular system into consideration in its bandwidth estimation method. Based on the review, we proposed the five tradeoff objectives to guide the CCA design in a mobile hybrid type of network: Bandwidth Utilisation, Delay (where loss is the extreme expression of delay), Fairness, Simplicity and Genericity. Based on the tradeoffs and goals, we proposed CDBE, a TCP clientside driven bandwidth estimation(CDBE) and report feedback loop. The client-side BW estimation method takes only little computation capability in the second version, compared to its first version and the DCIC. Cooperate with the enhanced server-side state transition CDBE can achieve a fair share of BW in both fixed packet core network or mobile network with a lower cost of RTT compared to conventional CCAs. No extra middlebox or edge computing unit/applications is required in CDBE architecture.
• Reactive or Stable: A Plant-inspired Approach for Business Organisation Morphogenesis
  
  • Zahada Payam
  • Diaconescu Ada
  , 2020.
• Real-time Anticipation of Occlusions for Automated Camera Control in Toric Space
  
  • Burg Ludovic
  • Lino Christophe
  • Christie Marc
  Computer Graphics Forum, Wiley, 2020, xx (2), pp.1 - 11. Efficient visibility computation is a prominent requirement when designing automated camera control techniques for dynamic3D environments; computer games, interactive storytelling or 3D media applications all need to track 3D entities while ensuringtheir visibility and delivering a smooth cinematic experience. Addressing this problem requires to sample a large set of potentialcamera positions and estimate visibility for each of them, which in practice is intractable despite the efficiency of ray-castingtechniques on recent platforms. In this work, we introduce a novel GPU-rendering technique to efficiently compute occlusionsof tracked targets in Toric Space coordinates – a parametric space designed for cinematic camera control. We then rely on thisocclusion evaluation to derive an anticipation map predicting occlusions for a continuous set of cameras over a user-definedtime window. We finally design a camera motion strategy exploiting this anticipation map to minimize the occlusions of trackedentities over time. The key features of our approach are demonstrated through comparison with traditionally used ray-castingon benchmark scenes, and through an integration in multiple game-like 3D scenes with heavy, sparse and dense occluders. (10.1111/cgf.13949)
  DOI : 10.1111/cgf.13949
• Duality in RKHSs with Infinite Dimensional Outputs: Application to Robust Losses
  
  • Laforgue Pierre
  • Lambert Alex
  • Brogat-Motte Luc
  • d'Alché-Buc Florence
  Proceedings of the 37th International Conference on Machine Learning, 2020. Operator-Valued Kernels (OVKs) and associated vector-valued Reproducing Kernel Hilbert Spaces provide an elegant way to extend scalar kernel methods when the output space is a Hilbert space. Although primarily used in finite dimension for problems like multi-task regression, the ability of this framework to deal with infinite dimensional output spaces unlocks many more applications, such as functional regression, structured output prediction, and structured data representation. However, these sophisticated schemes crucially rely on the kernel trick in the output space, so that most of previous works have focused on the square norm loss function, completely neglecting robustness issues that may arise in such surrogate problems. To overcome this limitation, this paper develops a duality approach that allows to solve OVK machines for a wide range of loss functions. The infinite dimensional Lagrange multipliers are handled through a Double Representer Theorem, and algorithms for-insensitive losses and the Huber loss are thoroughly detailed. Robustness benefits are emphasized by a theoretical stability analysis, as well as empirical improvements on structured data applications.
• Improved Optimistic Algorithms for Logistic Bandits
  
  • Faury Louis
  • Abeille Marc
  • Calauzènes Clément
  • Fercoq Olivier
  , 2020. The generalized linear bandit framework has attracted a lot of attention in recent years by extending the well-understood linear setting and allowing to model richer reward structures. It notably covers the logistic model, widely used when rewards are binary. For logistic bandits, the frequentist regret guarantees of existing algorithms areÕ(κ √ T), where κ is a problem-dependent constant. Unfortunately, κ can be arbitrarily large as it scales exponentially with the size of the decision set. This may lead to significantly loose regret bounds and poor empirical performance. In this work, we study the logistic bandit with a focus on the prohibitive dependencies introduced by κ. We propose a new optimistic algorithm based on a finer examination of the non-linearities of the reward function. We show that it enjoys aÕ(√ T) regret with no dependency in κ, but for a second order term. Our analysis is based on a new tail-inequality for self-normalized martingales, of independent interest.
• Contribution à des problèmes statistiques d'ordonnancement et d'apprentissage par renforcement avec aversion au risque
  
  • Achab Mastane
  , 2020. This thesis divides into two parts: the first part is on ranking and the second on risk-aware reinforcement learning. While binary classification is the flagship application of empirical risk minimization (ERM), the main paradigm of machine learning, more challenging problems such as bipartite ranking can also be expressed through that setup. In bipartite ranking, the goal is to order, by means of scoring methods, all the elements of some feature space based on a training dataset composed of feature vectors with their binary labels. This thesis extends this setting to the continuous ranking problem, a variant where the labels are taking continuous values instead of being simply binary. The analysis of ranking data, initiated in the 18th century in the context of elections, has led to another ranking problem using ERM, namely ranking aggregation and more precisely the Kemeny's consensus approach. From a training dataset made of ranking data, such as permutations or pairwise comparisons, the goal is to find the single "median permutation" that best corresponds to a consensus order. We present a less drastic dimensionality reduction approach where a distribution on rankings is approximated by a simpler distribution, which is not necessarily reduced to a Dirac mass as in ranking aggregation.For that purpose, we rely on mathematical tools from the theory of optimal transport such as Wasserstein metrics. The second part of this thesis focuses on risk-aware versions of the stochastic multi-armed bandit problem and of reinforcement learning (RL), where an agent is interacting with a dynamic environment by taking actions and receiving rewards, the objective being to maximize the total payoff. In particular, a novel atomic distributional RL approach is provided: the distribution of the total payoff is approximated by particles that correspond to trimmed means.
• Privacy-preserving content-based publish/subscribe with encrypted matching and data splitting
  
  • Denis Nathanaël
  • Chaffardon Pierre
  • Conan Denis
  • Laurent Maryline
  • Chabridon Sophie
  • Leneutre Jean
  , 2020, 3, pp.405-414. The content-based publish/subscribe paradigm enables a loosely-coupled and expressive form of communication. However, privacy preservation remains a challenge for distributed event-based middleware especially since encrypted matching incurs significant computing overhead. This paper adapts an existing attribute-based encryption scheme and combines it with data splitting, a non-cryptographic method called for alleviating the cost of encrypted matching. Data splitting enables to form groups of attributes that are sent apart over several independent broker networks so that it prevents the identification of an end-user; and, only identifying attributes are encrypted to prevent data leakage. The goal is to achieve an acceptable privacy level at an affordable computing price by encrypting only the necessary attributes, whose selection is determined through a Privacy Impact Assessment. (10.5220/0009833204050414)
  DOI : 10.5220/0009833204050414
• Validation platform for vehicle secure and highly trusted communications in the context of the cooperative ITS systems
  
  • Haidar Farah
  , 2020. Cooperative Intelligent Transportation System (C-ITS) has gained much attention in the recent years due to the large number of applications/use cases that can improve future driving experience. Future vehicles will be connected through several communication technologies which will open the door to new threats and vulnerabilities that must be taken into account. The security protection is a key subject to address before C-ITS deployment. Moreover, the wide variety of C-ITS use cases/application with different security requirements makes the security a big challenge. User's privacy and data protection are also a challenge. Automotive industry and operators should comply with the national and international legislation for the data protection in C-ITS. In order to deal with privacy issues, existing solution consists of having a pool of valid pseudonym identities, by the vehicle, and changing them during the communication. One of the motivations of this thesis is to study the performance of pseudonym certificate reloading. In other words, it is important to ensure that the latency of reloading pseudonym certificates from the PKI while driving at different speeds is acceptable. The second motivation is the investigation on threats and vulnerabilities, especially on those that come from the pseudonym certificate's use. The objective is to implement those attacks and propose new solutions or find improvements to the existing solution for detecting and preventing security attacks. Security and privacy in C-ITS are considered as big challenges. A Lot of work has been done and good solutions exist in the security and privacy domain. We notice that systems cannot be secure at 100% but driver's safety is related to system's security. For this, the aim of this thesis is to do white hack of the C-ITS in order to improve the existing solution. A risk assessment is needed to identify our target of evaluation and analyse potential risks. The final goal of this thesis is to propose a security and performance validation plate-form for vehicular communication in the context of C-ITS.
• User Association in Hybrid UAV-cellular Networks for Massive Real-time IoT Applications
  
  • Al Zahr Sawsan
  • Foroughi Parisa
  • Beyranvand Hamzeh
  • Gagnaire Maurice
  • Zahr Sawsan Al
  , 2020, pp.243-248. (10.1109/INFOCOMWKSHPS50562.2020.9162928)
  DOI : 10.1109/INFOCOMWKSHPS50562.2020.9162928
• ChainIDE 2.0: Facilitating Smart Contract Development for Consortium Blockchain
  
  • Wu Xiao
  • Qiu Han
  • Zhang Shuyi
  • Memmi Gerard
  • Gai Keke
  • Cai Wei
  , 2020, pp.388-393. (10.1109/INFOCOMWKSHPS50562.2020.9163051)
  DOI : 10.1109/INFOCOMWKSHPS50562.2020.9163051
• Reconfigurable Intelligent Surfaces: Bridging the gap between scattering and reflection
  
  • Garcia Juan Carlos Bucheli
  • Sibille Alain
  • Kamoun Mohamed
  IEEE Journal on Selected Areas in Communications, Institute of Electrical and Electronics Engineers, 2020, pp.1-1. (10.1109/JSAC.2020.3007037)
  DOI : 10.1109/JSAC.2020.3007037
• Influence of new network architectures and usages on RF human exposure in cellular networks
  
  • Chobineh Amirreza
  , 2020. In coming years, there will be a sharp growth in wireless data traffic due to the increasing usage of mobile phones and implementation of IoT technology. Therefore, mobile network operators aim to increase the capacity of their networks, to provide higher data traffic with lower latency, and to support thousands of connections. One of the primary efforts toward these goals is to densify today's cellular Macrocell networks using Small cells to bring more coverage and higher network capacity. Small cell antennas emit lower power than Macrocells and are often deployed at low heights. As a consequence, they are closer to the user and can be implemented massively. The latter can result in an important raise in public concerns. Mobile phones are used on the one hand, for a large variety of data usages that require different amounts of data and throughput and on the other hand for making phone calls. Voice over IP applications such as Skype has become very popular since they provide cheap international voice communication and can be used on mobile devices. Since LTE systems only support packet services, the voice service uses Voice over LTE technology instead of classical circuit-switched voice technology as in GSM and UMTS. The main objective of this thesis is to characterize and analyse the influence of new network architectures and usages on the actual human exposure induced by cellular networks. In this regard, several measurement campaigns were carried out in various cities and environments. Regarding the EMF exposure in heterogeneous networks, results suggest that by implementing Small cells, the global exposure (i.e. exposure induced by mobile phone and base station antenna) reduces due to the fact that by bringing the antenna closer to the user the emitted power by mobile phone and the usage duration reduce owing to power control schemes implemented in cellular network technologies. However, the magnitude of exposure reduction depends on the location of the Small cell with respect to Macrocells. Moreover, to assess the EMF exposure of indoor users induced by Small cells, two statistical models are proposed for the uplink and downlink exposures in an LTE heterogeneous environment based on measurements. The last part of the thesis was devoted to the assessment of the exposure for new types of usages through measurements. Results suggest that the amount of uplink emitted power and the emission time duration by a mobile phone is highly dependent on the usage and network technology. Voice call communications require a continuous and generally low throughput in order to maintain the communication during the call. On the contrary, in data usage, the mobile phone requires higher data and throughput to perform the task as fast as possible. Therefore during a voice call even if the user is using the mobile phone for a relatively long time, the exposure time duration should be lower since the usage does not require high amounts of data. The temporal occupation rate for several types of voice calls for different technologies is assessed through measurements.
• A Dynamic Clustering Algorithm for Multi-Point Transmissions in Mission-Critical Communications
  
  • Daher Alaa
  • Coupechoux Marceau
  • Godlewski Philippe
  • Ngouat Pierre
  • Minot Pierre
  IEEE Transactions on Wireless Communications, Institute of Electrical and Electronics Engineers, 2020, 19 (7), pp.4934-4946. Reliable group video call is one of the main services offered by future Mission-Critical Communications (MCC). To support its requirements, coordinated multi-point transmission in multi-cell environments is an attractive feature for MCC over Multimedia Broadcast Multicast Services owing to its potential for coverage improvement and multicast transmission. In such a scheme, full cooperation among all cells of an area achieves the highest cooperative gain, but has stringent impact on system capacity. A trade-off in the cluster's size of serving cells thus arises between high Signal to Interference plus Noise Ratio (SINR) and network capacity. In this paper, we formulate an optimization problem to maintain an acceptable system blocking probability, while maximizing the average SINR of the multicast group users. For every multicast group to be served, a dynamic cluster of cells is selected based on the minimization of a submodular function that takes into account the traffic in every cell through some weights and the average SINR achieved by the group users. Traffic weights are then optimized using a modified Nelder-Mead simplex method with the objective of tracking a blocking probability threshold. The proposed clustering scheme is compared to full cooperation and to Single-Cell Point-To-Multipoint (SC-PTM) schemes. Results show that dynamic clustering offers the best trade-off between coverage and capacity for MCC. (10.1109/TWC.2020.2988382)
  DOI : 10.1109/TWC.2020.2988382
• Actes de la 23e Conférence Nationale en Intelligence Artificielle
  
  • Bloch Isabelle
  , 2020.
• A Fully Stochastic Primal-Dual Algorithm
  
  • Bianchi Pascal
  • Hachem Walid
  • Salim Adil
  Optimization Letters, Springer Verlag, 2020. A new stochastic primal-dual algorithm for solving a composite optimization problem is proposed. It is assumed that all the functions / operators that enter the optimization problem are given as statistical expectations. These expectations are unknown but revealed across time through i.i.d realizations. The proposed algorithm is proven to converge to a saddle point of the Lagrangian function. In the framework of the monotone operator theory, the convergence proof relies on recent results on the stochastic Forward Backward algorithm involving random monotone operators. An example of convex optimization under stochastic linear constraints is considered. (10.1007/s11590-020-01614-y)
  DOI : 10.1007/s11590-020-01614-y
• Trisymmetric Multiplication Formulae in Finite Fields
  
  • Randriambololona Hugues
  • Rousseau Édouard
  , 2021, 12542, pp.92-111. Multiplication is an expensive arithmetic operation, therefore there has been extensive research to find Karatsuba-like formulae reducing the number of multiplications involved when computing a bilinear map. The minimal number of multiplications in such formulae is called the bilinear complexity, and it is also of theoretical interest to asymptotically understand it. Moreover, when the bilinear maps admit some kind of invariance, it is also desirable to find formulae keeping the same invariance. In this work, we study trisymmetric, hypersymmetric, and Galois invariant multiplication formulae over finite fields, and we give an algorithm to find such formulae. We also generalize the result that the bilinear complexity and symmetric bilinear complexity of the twovariable multiplication in an extension field are linear in the degree of the extension, to trisymmetric bilinear complexity, and to the complexity of t-variable multiplication for any t ≥ 3. (10.1007/978-3-030-68869-1_5)
  DOI : 10.1007/978-3-030-68869-1_5
• CSME: A novel cycle-sensing margin enhancement scheme for high yield STT-MRAM
  
  • Cai H.
  • Liu M.
  • Zhou Y.
  • Liu B.
  • Naviner Lirida
  Microelectronics Reliability, Elsevier, 2020, pp.113732. (10.1016/j.microrel.2020.113732)
  DOI : 10.1016/j.microrel.2020.113732
• Random extrapolation for primal-dual coordinate descent
  
  • Alacaoglu Ahmet
  • Fercoq Olivier
  • Cevher Volkan
  , 2020. We introduce a randomly extrapolated primal-dual coordinate descent method that adapts to sparsity of the data matrix and the favorable structures of the objective function. Our method updates only a subset of primal and dual variables with sparse data, and it uses large step sizes with dense data, retaining the benefits of the specific methods designed for each case. In addition to adapting to sparsity, our method attains fast convergence guarantees in favorable cases \textit{without any modifications}. In particular, we prove linear convergence under metric subregularity, which applies to strongly convex-strongly concave problems and piecewise linear quadratic functions. We show almost sure convergence of the sequence and optimal sublinear convergence rates for the primal-dual gap and objective values, in the general convex-concave case. Numerical evidence demonstrates the state-of-the-art empirical performance of our method in sparse and dense settings, matching and improving the existing methods.
• Distributed Hypothesis Testing based on Unequal-Error Protection Codes
  
  • Salehkalaibar Sadaf
  • Wigger Michèle
  IEEE Transactions on Information Theory, Institute of Electrical and Electronics Engineers, 2020. Coding and testing schemes for binary hypothesis testing over noisy networks are proposed and their corresponding type-II error exponents are derived. When communication is over a discrete memoryless channel (DMC), our scheme combines Shimokawa-Han-Amari’s hypothesis testing scheme with Borade-Nakiboglu-Zheng’s unequal error protection (UEP) for channel coding where source and channel codewords are simultaneously decoded. The resulting exponent is optimal for the newly introduced class of generalized testing against conditional independence. When communication is over a multi-access channel (MAC), our scheme combines hybrid coding with UEP. The resulting error exponent over the MAC is optimal in the case of generalized testing against conditional independence with independent observations at the two sensors when the MAC decomposes into two individual DMCs. In this case, separate source-channel coding is sufficient and no UEP is required. This same conclusion holds also under arbitrarily correlated sensor observations when testing is against independence. (10.1109/TIT.2020.2993172)
  DOI : 10.1109/TIT.2020.2993172