Publications

2020

• Magnetic Tunnel Junction Applications
  
  • Maciel Nilson
  • Marques Elaine
  • Naviner Lirida
  • Zhou Yongliang
  • Cai Hao
  Sensors, MDPI, 2020, 20 (1), pp.121. Spin-based devices can reduce energy leakage and thus increase energy efficiency. They have been seen as an approach to overcoming the constraints of CMOS downscaling, specifically, the Magnetic Tunnel Junction (MTJ) which has been the focus of much research in recent years. Its nonvolatility, scalability and low power consumption are highly attractive when applied in several components. This paper aims at providing a survey of a selection of MTJ applications such as memory and analog to digital converter, among others. (10.3390/s20010121)
  DOI : 10.3390/s20010121
• Towards Phase Balancing using Energy Storage
  
  • Hashmi Md Umar
  • Horta José
  • Pereira Lucas
  • Lee Zachary
  • Bušić Ana
  • Kofman Daniel
  , 2020. Ad-hoc growth of single-phase-connected distributed energy resources, such as solar generation and electric vehicles, can lead to network unbalance with negative consequences on the quality and efficiency of electricity supply. Case-studies are presented for a substation in Madeira, Portugal and an EV charging facility in Pasadena, California. These case studies show that phase imbalance can happen due to a large amount of distributed generation (DG) and electric vehicle (EV) integration. We conducted stylized load-flow analysis on a radial distribution network using an openDSS-based simulator to understand such negative effects of phase imbalance on neutral and phase conductor losses, and in voltage drop/rise. We evaluate the integration of storage in the distribution network as a possible solution for mitigating effects caused by imbalance. We present control architectures of storage operation for phase balancing. Numerically we show that relatively small-sized storage (compared to unbalance magnitude) can significantly reduce network imbalance. We identify the end node of the feeder as the best location to install storage. (10.48550/arXiv.2002.04177)
  DOI : 10.48550/arXiv.2002.04177
• Multimodal Analysis of Cohesion in Multi-party Interactions
  
  • B Kantharaju Reshmashree
  • Langlet Caroline
  • Barange Mukesh
  • Clavel Chloé
  • Pelachaud Catherine I
  , 2020. Group cohesion is an emergent phenomenon that describes the tendency of the group members' shared commitment to group tasks and the interpersonal attraction among them. This paper presents a multimodal analysis of group cohesion using a corpus of multi-party interactions. 16 two-minute segments annotated with cohesion data is used. We define three layers of modalities: non-verbal social cues, dialogue acts and interruptions. The initial analysis is performed at the individual level and later, we combine the different modalities to observe their impact on perceived level of cohesion. Results indicate that occurrence of laughter and interruption are higher in high cohesive segments. We also observed that, dialogue acts and head nods did not have an impact on the level of cohesion by itself. However, when combined there was an impact on the perceived level of cohesion. Overall, the analysis shows that multimodal cues are crucial for accurate analysis of group cohesion.
• Uniform convergence rates for the approximated halfspace and projection depth
  
  • Nagy Stanislav
  • Dyckerhoff Rainer
  • Mozharovskyi Pavlo
  Electronic Journal of Statistics, Shaker Heights, OH : Institute of Mathematical Statistics, 2020, 14 (2). (10.1214/20-EJS1759)
  DOI : 10.1214/20-EJS1759
• Scikit-network: Graph Analysis in Python
  
  • Bonald Thomas
  • de Lara Nathan
  • Lutz Quentin
  • Charpentier Bertrand
  Journal of Machine Learning Research, Microtome Publishing, 2020. Scikit-network is a Python package inspired by scikit-learn for the analysis of large graphs. Graphs are represented by their adjacency matrix in the sparse CSR format of SciPy. The package provides state-of-the-art algorithms for ranking, clustering, classifying, embedding and visualizing the nodes of a graph. High performance is achieved through a mix of fast matrix-vector products (using SciPy), compiled code (using Cython) and parallel processing. The package is distributed under the BSD license, with dependencies limited to NumPy and SciPy. It is compatible with Python 3.6 and newer. Source code, documentation and installation instructions are available online.
• A Model-Based Combination Language for Scheduling Verification
  
  • Zhao Hui
  • Apvrille Ludovic
  • Mallet Frédéric
  , 2020. Cyber-Physical Systems (CPSs) are built upon discrete software and hardware components, as well as continuous physical components. Such heterogeneous systems involve numerous domains with competencies and expertise that go far beyond traditional software engineering: systems engineering. In this paper , we explore a model-based approach for systems engineering that advocates the composition of several heterogeneous artifacts (called views) into a sound and consistent system model. A model combination Language is proposed for this purpose. Thus, rather than trying to build the universal language able to capture all possible aspects of systems, the proposed language proposes to relate small subsets of languages in order to offer specific analysis capabilities while keeping a global consistency between all joined models. We demonstrate the interest of our approach through an industrial process based on Capella, which provides (among others) a large support for functional analysis from requirements to components deployment. Even though Capella is already quite expressive, it lacks support for schedulability analysis. AADL is also a language dedicated to system analysis. If it is backed with advanced schedulability tools, it lacks support for functional analysis. Thus, instead of proposing ways to add missing aspects in either Capella or AADL, we rather extract a relevant subset of both languages to build a view adequate for conducting schedulability analysis of Capella functional models. Finally, our combination language is generic enough to extract pertinent subsets of languages and combine them to build views for different experts. It also helps maintaining a global consistency between different modeling views.
• Monadic Datalog, Tree Validity, and Limited Access Containment
  
  • Benedikt Michael
  • Bourhis Pierre
  • Gottlob Georg
  • Senellart Pierre
  ACM Transactions on Computational Logic, Association for Computing Machinery, 2020, 21 (1), pp.6:1-6:45. We reconsider the problem of containment of monadic datalog (MDL) queries in unions of conjunctive queries (UCQs). Prior work has dealt with special cases of the problem, but has left the precise complexity characterization open. In addition, the complexity of one important special case, that of containment under access patterns, was not known before. We start by revisiting the connection between MDL/UCQ containment and containment problems involving regular tree languages. We then present a general approach for getting tighter bounds on the complexity of query containment, based on analysis of the number of mappings of queries into tree-like instances. We give two applications of the machinery. We first give an important special case of the MDL/UCQ containment problem that is in EXPTIME, and use this bound to show an EXPTIME bound on containment under access patterns. Secondly we show that the same technique can be used to get a new tight upper bound for containment of tree automata in UCQs. We finally show that the new MDL/UCQ upper bounds are tight. We establish a 2EXPTIME lower bound on the MDL/UCQ containment problem, resolving an open problem from the early 1990s. This bound holds for the MDL/CQ containment problem as well. We also show that changes to the conditions given in our special cases can not be eliminated, and that in particular slight variations of the problem of containment under access patterns become 2EXPTIME-complete. (10.1145/3344514)
  DOI : 10.1145/3344514
• SPECTRAL EMBEDDING OF REGULARIZED BLOCK MODELS
  
  • de Lara Nathan
  • Bonald Thomas
  , 2020. Spectral embedding is a popular technique for the representation of graph data. Several regularization techniques have been proposed to improve the quality of the embedding with respect to downstream tasks like clustering. In this paper, we explain on a simple block model the impact of the complete graph regularization, whereby a constant is added to all entries of the adjacency matrix. Specifically, we show that the regularization forces the spectral embedding to focus on the largest blocks, making the representation less sensitive to noise or outliers. We illustrate these results on both on both synthetic and real data, showing how regularization improves standard clustering scores.
• Template-Based Graph Clustering
  
  • Riva M.
  • Yger F.
  • Gori P.
  • Cesar R.
  • Bloch Isabelle
  , 2020.
• Quantitative Propagation of Chaos for SGD in Wide Neural Networks
  
  • de Bortoli Valentin
  • Durmus Alain
  • Fontaine Xavier
  • Şimşekli Umut
  , 2020. In this paper, we investigate the limiting behavior of a continuous-time counterpart of the Stochastic Gradient Descent (SGD) algorithm applied to two-layer overparameterized neural networks, as the number or neurons (ie, the size of the hidden layer) $N \to +\infty$. Following a probabilistic approach, we show 'propagation of chaos' for the particle system defined by this continuous-time dynamics under different scenarios, indicating that the statistical interaction between the particles asymptotically vanishes. In particular, we establish quantitative convergence with respect to $N$ of any particle to a solution of a mean-field McKean-Vlasov equation in the metric space endowed with the Wasserstein distance. In comparison to previous works on the subject, we consider settings in which the sequence of stepsizes in SGD can potentially depend on the number of neurons and the iterations. We then identify two regimes under which different mean-field limits are obtained, one of them corresponding to an implicitly regularized version of the minimization problem at hand. We perform various experiments on real datasets to validate our theoretical results, assessing the existence of these two regimes on classification problems and illustrating our convergence results.
• Statistical and Topological Properties of Sliced Probability Divergences
  
  • Nadjahi Kimia
  • Durmus Alain
  • Chizat Lénaïc
  • Kolouri Soheil
  • Shahrampour Shahin
  • Şimşekli Umut
  , 2020. The idea of slicing divergences has been proven to be successful when comparing two probability measures in various machine learning applications including generative modeling, and consists in computing the expected value of a `base divergence' between one-dimensional random projections of the two measures. However, the computational and statistical consequences of such a technique have not yet been well-established. In this paper, we aim at bridging this gap and derive some properties of sliced divergence functions. First, we show that slicing preserves the metric axioms and the weak continuity of the divergence, implying that the sliced divergence will share similar topological properties. We then precise the results in the case where the base divergence belongs to the class of integral probability metrics. On the other hand, we establish that, under mild conditions, the sample complexity of the sliced divergence does not depend on the dimension, even when the base divergence suffers from the curse of dimensionality. We finally apply our general results to the Wasserstein distance and Sinkhorn divergences, and illustrate our theory on both synthetic and real data experiments.
• Hausdorff Dimension, Heavy Tails, and Generalization in Neural Networks
  
  • Şimşekli Umut
  • Sener Ozan
  • Deligiannidis George
  • Erdogdu Murat A.
  , 2020. Despite its success in a wide range of applications, characterizing the generalization properties of stochastic gradient descent (SGD) in non-convex deep learning problems is still an important challenge. While modeling the trajectories of SGD via stochastic differential equations (SDE) under heavy-tailed gradient noise has recently shed light over several peculiar characteristics of SGD, a rigorous treatment of the generalization properties of such SDEs in a learning theoretical framework is still missing. Aiming to bridge this gap, in this paper, we prove generalization bounds for SGD under the assumption that its trajectories can be well-approximated by a \emph{Feller process}, which defines a rich class of Markov processes that include several recent SDE representations (both Brownian or heavy-tailed) as its special case. We show that the generalization error can be controlled by the \emph{Hausdorff dimension} of the trajectories, which is intimately linked to the tail behavior of the driving process. Our results imply that heavier-tailed processes should achieve better generalization; hence, the tail-index of the process can be used as a notion of "capacity metric". We support our theory with experiments on deep neural networks illustrating that the proposed capacity metric accurately estimates the generalization error, and it does not necessarily grow with the number of parameters unlike the existing capacity metrics in the literature.
• Stein's method for diffusive limit of queueing processes
  
  • Besançon Eustache
  • Decreusefond Laurent
  • Moyal Pascal
  Queueing Systems, Springer Verlag, 2020, 95, pp.173--201. Donsker Theorem is perhaps the most famous invariance principle result for Markov processes. It states that when properly normalized, a random walk behaves asymptotically like a Brownian motion. This approach can be extended to general Markov processes whose driving parameters are taken to a limit, which can lead to insightful results in contexts like large distributed systems or queueing networks. The purpose of this paper is to assess the rate of convergence in these so-called diffusion approximations, in a queueing context. To this end, we extend the functional Stein method introduced for the Brownian approximation of Poisson processes, to two simple examples: the single-server queue and the infinite-server queue. By doing so, we complete the recent applications of Stein's method to queueing systems, with results concerning the whole trajectory of the considered process, rather than its stationary distribution. (10.1007/s11134-020-09658-8)
  DOI : 10.1007/s11134-020-09658-8
• Fidelity metrics between curves and surfaces: currents, varifolds, and normal cycles
  
  • Charon Nicolas
  • Charlier Benjamin
  • Glaunès Joan
  • Gori Pietro
  • Roussillon Pierre
  , 2020, pp.441-477. This chapter provides an overview of some mathematical and computational models that have been proposed over the past few years for defining data attachment terms on shape spaces of curves or surfaces. In all these models shapes are seen as elements of a space of generalized distributions, such as currents or varifolds. Then norms are defined through reproducing kernel Hilbert spaces (RKHS), which lead to shape distances that can be conveniently computed in practice. These were originally introduced in conjunction with diffeomorphic methods in computational anatomy and have indeed proved to be very efficient in this field. We provide a basic description of these different models and their practical implementation, then discuss the respective properties and potential advantages or downsides of each of them in diffeomorphic registration problems. (10.1016/B978-0-12-814725-2.00021-2)
  DOI : 10.1016/B978-0-12-814725-2.00021-2
• Introducing coherent MIMO sensing, a fading-resilient, polarization-independent approach to ϕ-OTDR
  
  • Guerrier Sterenn
  • Dorize Christian
  • Awwad Elie
  • Renaudier Jeremie
  Optics Express, Optical Society of America - OSA Publishing, 2020, 28 (14), pp.21081. (10.1364/OE.396460)
  DOI : 10.1364/OE.396460
• Storage-Computation-Communication Tradeoff in Distributed Computing: Fundamental Limits and Complexity
  
  • Yan Qifa
  • Yang Sheng
  • Wigger Michèle
  IEEE Transactions on Information Theory, Institute of Electrical and Electronics Engineers, 2020. Distributed computing has become one of the most important frameworks in dealing with large computation tasks. In this paper, we propose a systematic construction of coded computing schemes for MapReduce-type distributed systems. The construction builds upon placement delivery arrays (PDA), originally proposed by Yan et al. for coded caching schemes. The main contributions of our work are threefold. First, we identify a class of PDAs, called Comp-PDAs, and show how to obtain a coded computing scheme from any Comp-PDA. We also characterize the normalized number of stored files (storage load), computed intermediate values (computation load), and communicated bits (communication load), of the obtained schemes in terms of the Comp-PDA parameters. Then, we show that the performance achieved by Comp-PDAs describing Maddah-Ali and Niesen's coded caching schemes matches a new information-theoretic converse, thus establishing the fundamental region of all achievable performance triples. In particular, we characterize all the Comp-PDAs achieving the pareto-optimal storage, computation, and communication (SCC) loads of the fundamental region. Finally, we investigate the file complexity of the proposed schemes, i.e., the smallest number of files required for implementation. In particular, we describe Comp-PDAs that achieve pareto-optimal SCC triples with significantly lower file complexity than the originally proposed Comp-PDAs.
• Combined Vector Perturbation for Adaptive Modulation in MU-MIMO
  
  • Askri Aymen
  • Rekaya-Ben Othman Ghaya
  , 2020.
• Degrees-of-Freedom in Multi-Cloud Based Sectored Cellular Networks
  
  • Gelincik Samet
  • Rekaya-Ben Othman Ghaya
  Entropy, MDPI, 2020.
• An analysis of the transfer learning of convolutional neural networks for artistic images
  
  • Gonthier Nicolas
  • Gousseau Yann
  • Ladjal Saïd
  , 2020. Transfer learning from huge natural image datasets, fine-tuning of deep neural networks and the use of the corresponding pre-trained networks have become de facto the core of art analysis applications. Nevertheless, the effects of transfer learning are still poorly understood. In this paper, we first use techniques for visualizing the network internal representations in order to provide clues to the understanding of what the network has learned on artistic images. Then, we provide a quantitative analysis of the changes introduced by the learning process thanks to metrics in both the feature and parameter spaces, as well as metrics computed on the set of maximal activation images. These analyses are performed on several variations of the transfer learning procedure. In particular, we observed that the network could specialize some pre-trained filters to the new image modality and also that higher layers tend to concentrate classes. Finally, we have shown that a double fine-tuning involving a medium-size artistic dataset can improve the classification on smaller datasets, even when the task changes. (10.1007/978-3-030-68796-0_39)
  DOI : 10.1007/978-3-030-68796-0_39
• Uncovering reflection insensitive semiconductor lasers for silicon photonic integration (invited)
  
  • Grillot Frédéric
  , 2020, pp.M4H.4. (10.1364/OFC.2020.M4H.4)
  DOI : 10.1364/OFC.2020.M4H.4
• Solving $x+x^{2^l}+\cdots+x^{2^{ml}}=a$ over $\mathbb{F}_{2^n}$.
  
  • Mesnager Sihem
  • Kim K.H.
  • Choe J.H.
  • Lee D.N.
  • Go D.S.
  Cryptography and Communications-Discrete Structures, Boolean Functions, and Sequences (CCDS), 2020.
• On the boomerang uniformity of quadratic permutations
  
  • Mesnager Sihem
  • Tang C.
  • Xiong M.
  Journal of Designs, Codes, and Cryptography, 2020.
• The Need to Move beyond Triples
  
  • Suchanek Fabian
  , 2020. Almost all major knowledge bases are concerned mainly with binary relationships between entities. In this vision paper, we argue that it is time to broaden this view: first to relations of higher arity, complex objects, and events, and then also to knowledge about knowledge: We should be able to represent why something is true, that something is not true, that something happened before something else, or that something is mainly believed. While this idea is as old as Artificial Intelligence itself, we argue that only now we have the tools to achieve it: a better understanding of our use-cases and large amounts of data. We survey relevant approaches, and point out avenues of research.
• Advances in Neural Information Processing Systems 32 (NeurIPS 2019)
  
  • Wallach Hanna M.
  • Larochelle Hugo
  • Beygelzimer Alina
  • d'Alché-Buc Florence
  • Fox Emily B.
  • Garnett Roman
  , 2020.
• Space-Time Coding Performance Analysis for CDL-impaired Multi-Core Fiber Transmission
  
  • Abouseif Akram
  • Rekaya-Ben Othman Ghaya
  • Jaouën Yves
  , 2020.