Publications

2022

• Lyrics segmentation via bimodal text–audio representation
  
  • Fell Michael
  • Nechaev Yaroslav
  • Meseguer-Brocal Gabriel
  • Cabrio Elena
  • Gandon Fabien
  • Peeters Geoffroy
  Natural Language Engineering, Cambridge University Press (CUP), 2022, 28 (3), pp.317 - 336. Song lyrics contain repeated patterns that have been proven to facilitate automated lyrics segmentation, with the final goal of detecting the building blocks (e.g., chorus, verse) of a song text. Our contribution in this article is twofold. First, we introduce a convolutional neural network (CNN)-based model that learns to segment the lyrics based on their repetitive text structure. We experiment with novel features to reveal different kinds of repetitions in the lyrics, for instance based on phonetical and syntactical properties. Second, using a novel corpus where the song text is synchronized to the audio of the song, we show that the text and audio modalities capture complementary structure of the lyrics and that combining both is beneficial for lyrics segmentation performance. For the purely text-based lyrics segmentation on a dataset of 103k lyrics, we achieve an F-score of 67.4%, improving on the state of the art (59.2% F-score). On the synchronized text–audio dataset of 4.8k songs, we show that the additional audio features improve segmentation performance to 75.3% F-score, significantly outperforming the purely text-based approaches. (10.1017/S1351324921000024)
  DOI : 10.1017/S1351324921000024
• Uniform Reliability of Self-Join-Free Conjunctive Queries
  
  • Amarilli Antoine
  • Kimelfeld Benny
  Logical Methods in Computer Science, Logical Methods in Computer Science Association, 2022. The reliability of a Boolean Conjunctive Query (CQ) over a tuple-independent probabilistic database is the probability that the CQ is satisfied when the tuples of the database are sampled one by one, independently, with their associated probability. For queries without self-joins (repeated relation symbols), the data complexity of this problem is fully characterized by a known dichotomy: reliability can be computed in polynomial time for hierarchical queries, and is #P-hard for non-hierarchical queries. Inspired by this dichotomy, we investigate a fundamental counting problem for CQs without self-joins: how many sets of facts from the input database satisfy the query? This is equivalent to the uniform case of the query reliability problem, where the probability of every tuple is required to be 1/2. Of course, for hierarchical queries, uniform reliability is solvable in polynomial time, like the reliability problem. We show that being hierarchical is also necessary for this tractability (under conventional complexity assumptions). In fact, we establish a generalization of the dichotomy that covers every restricted case of reliability in which the probabilities of tuples are determined by their relation. (10.46298/lmcs-18(4:3)2022)
  DOI : 10.46298/lmcs-18(4:3)2022
• Reasoning about Moving Target Defense in Attack Modeling Formalisms
  
  • Ballot Gabriel
  • Malvone Vadim
  • Leneutre Jean
  • Borde Etienne
  , 2022. Since 2009, Moving Target Defense (MTD) has become a new paradigm of defensive mechanism that frequently changes the state of the target system to confuse the attacker. This frequent change is costly and leads to a trade-off between misleading the attacker and disrupting the quality of service. Optimizing the MTD activation frequency is necessary to develop this defense mechanism when facing realistic, multi-step attack scenarios. Attack modeling formalisms based on DAG are prominently used to specify these scenarios. Our contribution is a new DAG-based formalism for MTDs and its translation into a Price Timed Markov Decision Process to find the best activation frequencies against the attacker's time/cost-optimal strategies. For the first time, MTD activation frequencies are analyzed in a state-of-the-art DAG-based representation. Moreover, this is the first paper that considers the specificity of MTDs in the automatic analysis of attack modeling formalisms. Finally, we present some experimental results using Uppaal Stratego to demonstrate its applicability and relevance.
• (Complex) natural gradient optimization for optical quantum circuit design
  
  • Yao Yuan
  • Cussenot Pierre
  • Wolf Richard A.
  • Miatto Filippo M.
  Physical Review A, American Physical Society, 2022, 105 (5), pp.052402. Few-mode optical quantum circuits can be simulated and optimized using gradient-descent methods, as optical gates can be parametrized by continuous parameters. However, the parameter space as seen by a nonholomorphic cost function is not Euclidean, which means that the Euclidean gradient does not generally point in the direction of steepest ascent. In order to retrieve the true steepest-ascent direction, one must take into account the local metric in what is known as the natural-gradient (NG) method. In this work we implement the NG for the optimization of optical quantum circuits. Specifically, we adapt the NG approach to a complex-valued parameter space, allowing us to apply the NG method directly within the Bargmann formalism, which we rely on for transforming between phase space and Fock space. We then compare the NG approach to vanilla gradient descent and to Adam over two standard state-preparation benchmarks: a single-photon source and a Gottesman-Kitaev-Preskill state source. We observe that the NG approach converges significantly faster (due in part to the possibility of using larger learning rates) and with a smoother decay of the cost function throughout the optimization. This result further improves our ability to design quantum circuits via classical optimization methods. (10.1103/PhysRevA.105.052402)
  DOI : 10.1103/PhysRevA.105.052402
• Introduction
  
  • Bloch Isabelle
  • Euzenat Jérôme
  • Lang Jérôme
  • Schwarzentruber François
  Revue Ouverte d'Intelligence Artificielle, Association pour la diffusion de la recherche francophone en intelligence artificielle, 2022, 3, pp.193-199. no abstract (10.5802/roia.28fr)
  DOI : 10.5802/roia.28fr
• A Quadrature Rule combining Control Variates and Adaptive Importance Sampling
  
  • Leluc Rémi
  • Portier François
  • Zhuman Aigerim
  • Segers Johan
  Advances in Neural Information Processing Systems, Morgan Kaufmann Publishers, 2022, 35, pp.11842--11853. Driven by several successful applications such as in stochastic gradient descent or in Bayesian computation, control variates have become a major tool for Monte Carlo integration. However, standard methods do not allow the distribution of the particles to evolve during the algorithm, as is the case in sequential simulation methods. Within the standard adaptive importance sampling framework, a simple weighted least squares approach is proposed to improve the procedure with control variates. The procedure takes the form of a quadrature rule with adapted quadrature weights to reflect the information brought in by the control variates. The quadrature points and weights do not depend on the integrand, a computational advantage in case of multiple integrands. Moreover, the target density needs to be known only up to a multiplicative constant. Our main result is a non-asymptotic bound on the probabilistic error of the procedure. The bound proves that for improving the estimate's accuracy, the benefits from adaptive importance sampling and control variates can be combined. The good behavior of the method is illustrated empirically on synthetic examples and real-world data for Bayesian linear regression.
• Handwriting and Drawing Features for Detecting Personality Traits: An Analysis on Big Five Sub-dimensions
  
  • Esposito Anna
  • Amorese Terry
  • Buonanno Michele
  • Cuciniello Marialucia
  • Esposito Antonietta
  • Faundez-Zanuy Marcos
  • Likforman-Sulem Laurence
  • Riviello Maria Teresa
  • Spagnuolo Carmine
  • Troncone Alda
  • Cordasco Gennaro
  Acta Polytechnica Hungarica, Óbuda University, 2022, 19 (11), pp.65-84. : Handwriting and Drawing are functional tasks involving physical and cognitive processes. Recently they have been investigated for detecting cognitive and motor disorders. In this work, handwriting/drawing features are investigated for identifying connections with personality traits. For this purpose, an experiment comprising seven handwriting/drawing tasks has been administrated to 78 young adults (mean age=24.6 ± 2.4 years) equally balanced by gender. Handwriting and Drawing activities - both on and close to the paper – had been recorded online through a digitizing tablet able to measure handwriting and drawing features such as pressure, speed, dimension, and inclination of each pen-stroke on the paper. Participants were asked to fill the Big Five Personality Questionnaire (BFQ) and according to the scores obtained for each of the 5 dimensions and 10 Big Five sub-dimensions, were partitioned into three categories: low, typical, and high. To evaluate whether the recorded handwriting/drawing features are connected with personality traits ANOVA repeated measures have been performed with gender and group category (low, typical, and high) as between and the listed handwriting/drawing features as within factors. The analyses show significant differences among low, typical and, high BFQ scores for the main Big Five dimensions and the ten Big Five sub-dimensions, indicating that personality traits can be revealed by a quantitative analysis of the proposed handwriting/drawing features. (10.12700/APH.19.11.2022.11.4)
  DOI : 10.12700/APH.19.11.2022.11.4
• Selected Topics in Malliavin Calculus
  
  • Decreusefond Laurent
  , 2022, 10. (10.1007/978-3-031-01311-9)
  DOI : 10.1007/978-3-031-01311-9
• Enhanced four-wave mixing dynamics in epitaxial quantum dot laser on silicon
  
  • Ding Shihao
  • Dong Bozhang
  • Chow Weng
  • Bowers John
  • Grillot Frédéric
  , 2022, pp.NpTh3D.1. The four-wave mixing conversion efficiency of quantum dot laser is much higher than that of quantum well. These results are important for self-mode-locked pulse production and high-bandwidth optical frequency comb generation. (10.1364/NP.2022.NpTh3D.1)
  DOI : 10.1364/NP.2022.NpTh3D.1
• High-capacity free-space optical link in the midinfrared thermal atmospheric windows using unipolar quantum devices
  
  • Didier Pierre
  • Dely Hamza
  • Bonazzi Thomas
  • Spitz Olivier
  • Awwad Elie
  • Rodriguez Etienne
  • Vasanelli Angela
  • Sirtori Carlo
  • Grillot Frédéric
  Advanced photonics, SPIE, 2022, 4 (5), pp.056004. Free-space optical communication is a very promising alternative to fiber communication systems, in terms of ease of deployment and costs. Midinfrared light has several features of utter relevance for free-space applications: low absorption when propagating in the atmosphere even under adverse conditions, robustness of the wavefront during long-distance propagation, and absence of regulations and restrictions for this range of wavelengths. A proof-of-concept of high-speed transmission taking advantage of intersubband devices has recently been demonstrated, but this effort was limited by the short-distance optical path (up to 1 m). In this work, we study the possibility of building a long-range link using unipolar quantum optoelectronics. Two different detectors are used: an uncooled quantum cascade detector and a nitrogen-cooled quantum well-infrared photodetector. We evaluate the maximum data rate of our link in a back-to-back configuration before adding a Herriott cell to increase the length of the light path up to 31 m. By using pulse shaping, pre- and post-processing, we reach a record bitrate of 30 Gbit s − 1 for both two-level (OOK) and four-level (PAM-4) modulation schemes for a 31-m propagation link and a bit error rate compatible with error-correction codes. (10.1117/1.AP.4.5.056004)
  DOI : 10.1117/1.AP.4.5.056004
• Compositional Equivalences Based on Open pNets
  
  • Ameur-Boulifa Rabéa
  • Henrio Ludovic
  • Madelaine Eric
  Journal of Logical and Algebraic Methods in Programming, Elsevier, 2022, 131, pp.100842. Establishing equivalences between programs is crucial both for verifying correctness of programs and for justifying optimisations and program transformations. There exist several equivalence relations for programs, and bisimulations are among the most versatile of these equivalences. Among bisimulations one distinguishes strong bisimulation that requires that each action of a program is simulated by a single action of the equivalent program, and weak bisimulation that allows some of the actions to be invisible, and thus not simulated. pNet is a generalisation of automata that model open systems. They feature variables and hierarchical composition. Open pNets are pNets with holes, i.e. placeholders that can be filled later by subsystems. However, there is no standard tool for defining the semantics of an open system in this context. This article first defines open automata that are labelled transition systems with parameters and holes. Relying on open automata, it then defines bisimilarity relations for the comparison of systems specified as pNets. We first present a strong bisimilarity for open pNets called FH-bisimilarity. Next we offer an equivalence relation similar to the classical weak bisimulation equivalence, and study its properties. Among these properties we are interested in compositionality: if two systems are proven equivalent they will be indistinguishable by their context, and they will also be indistinguishable when their holes are filled with equivalent systems. We identify sufficient conditions to ensure compositionality of strong and weak bisimulation. The contributions of this article are illustrated using a transport protocol as running example. (10.1016/j.jlamp.2022.100842)
  DOI : 10.1016/j.jlamp.2022.100842
• LAKE DETECTION WITH SENTINEL-1 DATA USING A GRAB-CUT METHOD AND ITS MULTI-TEMPORAL EXTENSION
  
  • Gasnier Nicolas
  • Denis Loïc
  • Fjørtoft Roger
  • Liege Frédéric
  • Tupin Florence
  , 2022. This paper presents a semi-guided method to detect lakes in Sentinel-1 SAR data. The proposed approach is an adaptation of the grab-cut framework developed in [1]. Starting from a coarse bounding box around the lake, an accurate segmentation is extracted using a Conditional Random Field formalism and a graph-cut based optimization. Then an extension of this approach to process jointly a stack of multi-temporal data is presented. A temporal regularization term is introduced to control the joint segmentation. The proposed approach is evaluated on Sentinel-1 datasets. Qualitative and quantitative results demonstrate the interest of the proposed framework and its robustness to the initialization polygon of the lake. (10.1109/IGARSS46834.2022.9883219)
  DOI : 10.1109/IGARSS46834.2022.9883219
• Post-actes de la Conférence Nationale en Intelligence Artificielle (CNIA 2018-2020)
  
  • Bloch Isabelle
  • Euzenat Jérôme
  • Lang Jérôme
  • Schwarzentruber François
  Revue Ouverte d'Intelligence Artificielle, Association pour la diffusion de la recherche francophone en intelligence artificielle, 2022, 3, pp.193-413. no abstract
• Combining Embeddings and Rules for Fact Prediction
  
  • Boschin Armand
  • Jain Nitisha
  • Keretchashvili Gurami
  • Suchanek Fabian M.
  , 2022. Knowledge bases are typically incomplete, meaning that they are missing information that we would expect to be there. Recent years have seen two main approaches to guess missing facts: Rule Mining and Knowledge Graph Embeddings. The first approach is symbolic, and finds rules such as "If two people are married, they most likely live in the same city". These rules can then be used to predict missing statements. Knowledge Graph Embeddings, on the other hand, are trained to predict missing facts for a knowledge base by mapping entities to a vector space. Each of these approaches has their strengths and weaknesses, and this article provides a survey of neuro-symbolic works that combine embeddings and rule mining approaches for fact prediction. (10.4230/OASIcs.AIB.2022.4)
  DOI : 10.4230/OASIcs.AIB.2022.4
• Unsupervised Audio Source Separation Using Differentiable Parametric Source Models
  
  • Schulze-Forster Kilian
  • Doire Clement S J
  • Richard Gael
  • Badeau Roland
  Computing Research Repository, ACM / ArXiv, 2022. Supervised deep learning approaches to underdetermined audio source separation achieve state-of-the-art performance but require a dataset of mixtures along with their corresponding isolated source signals. Such datasets can be extremely costly to obtain for musical mixtures. This raises a need for unsupervised methods. We propose a novel unsupervised modelbased deep learning approach to musical source separation. Each source is modelled with a differentiable parametric sourcefilter model. A neural network is trained to reconstruct the observed mixture as a sum of the sources by estimating the source models' parameters given their fundamental frequencies. At test time, soft masks are obtained from the synthesized source signals. The experimental evaluation on a vocal ensemble separation task shows that the proposed method outperforms learning-free methods based on nonnegative matrix factorization and a supervised deep learning baseline. Integrating domain knowledge in the form of source models into a data-driven method leads to high data efficiency: the proposed approach achieves good separation quality even when trained on less than three minutes of audio. This work makes powerful deep learning based separation usable in scenarios where training data with ground truth is expensive or nonexistent.
• Characterization of the majority matrices of profiles of equivalence relations
  
  • Hudry Olivier
  , 2022.
• From Coherent Systems Technology to Advanced Fiber Sensing for Smart Network Monitoring
  
  • Dorize Christian
  • Guerrier Sterenn
  • Awwad Elie
  • Mardoyan Haik
  • Renaudier Jeremie
  Journal of Lightwave Technology, Institute of Electrical and Electronics Engineers (IEEE)/Optical Society of America(OSA), 2022, pp.1-10. (10.1109/JLT.2022.3221552)
  DOI : 10.1109/JLT.2022.3221552
• Towards Globally Optimized Hybrid Homomorphic Encryption - Featuring the Elisabeth Stream Cipher
  
  • Cosseron Orel
  • Hoffmann Clément
  • Méaux Pierrick
  • Standaert François-Xavier
  , 2022.
• Listen to Interpret: Post-hoc Interpretability for Audio Networks with NMF
  
  • Jayneel Parekh
  • Sanjeel Parekh
  • Pavlo Mozharovskyi
  • Florence d'Alché-Buc
  • Richard Gael
  , 2022. This paper tackles post-hoc interpretability for audio processing networks. Our goal is to interpret decisions of a trained network in terms of high-level audio objects that are also listenable for the end-user. To this end, we propose a novel interpreter design that incorporates non-negative matrix factorization (NMF). In particular, a regularized interpreter module is trained to take hidden layer representations of the targeted network as input and produce time activations of pre-learnt NMF components as intermediate outputs. Our methodology allows us to generate intuitive audio-based interpretations that explicitly enhance parts of the input signal most relevant for a network's decision. We demonstrate our method's applicability on popular benchmarks, including a real-world multi-label classification task.
• DNN-FREE LOW-LATENCY ADAPTIVE SPEECH ENHANCEMENT BASED ON FRAME-ONLINE BEAMFORMING POWERED BY BLOCK-ONLINE FASTMNMF
  
  • Nugraha Aditya Arie
  • Sekiguchi Kouhei
  • Fontaine Mathieu
  • Bando Yoshiaki
  • Yoshii Kazuyoshi
  , 2022. This paper describes a practical dual-process speech enhancement system that adapts environment-sensitive frame-online beamforming (front-end) with help from environment-free block-online source separation (back-end). To use minimum variance distortionless response (MVDR) beamforming, one may train a deep neural network (DNN) that estimates timefrequency masks used for computing the covariance matrices of sources (speech and noise). Backpropagation-based runtime adaptation of the DNN was proposed for dealing with the mismatched training-test conditions. Instead, one may try to directly estimate the source covariance matrices with a state-ofthe-art blind source separation method called fast multichannel non-negative matrix factorization (FastMNMF). In practice, however, neither the DNN nor the FastMNMF can be updated in a frame-online manner due to its computationally-expensive iterative nature. Our DNN-free system leverages the posteriors of the latest source spectrograms given by block-online FastMNMF to derive the current source covariance matrices for frame-online beamforming. The evaluation shows that our frame-online system can quickly respond to scene changes caused by interfering speaker movements and outperformed an existing block-online system with DNN-based beamforming by 5.0 points in terms of the word error rate.
• Statistical learning from biased training samples
  
  • Clémençon Stéphan
  • Laforgue Pierre
  Electronic Journal of Statistics, Shaker Heights, OH : Institute of Mathematical Statistics, 2022, 16 (2), pp.6086-6134. With the deluge of digitized information in the Big Data era, massive datasets are becoming increasingly available for learning predictive models. However, in many practical situations, the poor control of the data acquisition processes may naturally jeopardize the outputs of machine learning algorithms, and selection bias issues are now the subject of much attention in the literature. The present article investigates how to extend Empirical Risk Minimization, the principal paradigm in statistical learning, when training observations are generated from biased models, i.e., from distributions that are different from that in the test/prediction stage, and absolutely continuous with respect to the latter. Precisely, we show how to build a “nearly debiased” training statistical population from biased samples and the related biasing functions, following in the footsteps of the approach originally proposed in [46]. Furthermore, we study from a nonasymptotic perspective the performance of minimizers of an empirical version of the risk computed from the statistical population thus created. Remarkably, the learning rate achieved by this procedure is of the same order as that attained in absence of selection bias. Beyond the theoretical guarantees, we also present experimental results supporting the relevance of the algorithmic approach promoted in this paper. (10.1214/22-EJS2084)
  DOI : 10.1214/22-EJS2084
• Cybersecurity in Smart Homes: Architectures, Solutions and Technologies
  
  • Khatoun Rida
  , 2022. Smart homes use Internet-connected devices, artificial intelligence, protocols and numerous technologies to enable people to remotely monitor their home, as well as manage various systems within it via the Internet using a smartphone or a computer. A smart home is programmed to act autonomously to improve comfort levels, save energy and potentially ensure safety; the result is a better way of life. Innovative solutions continue to be developed by researchers and engineers and thus smart home technologies are constantly evolving. By the same token, cybercrime is also becoming more prevalent. Indeed, a smart home system is made up of connected devices that cybercriminals can infiltrate to access private information, commit cyber vandalism or infect devices using botnets. This book addresses cyber attacks such as sniffing, port scanning, address spoofing, session hijacking, ransomware and denial of service. It presents, analyzes and discusses the various aspects of cybersecurity as well as solutions proposed by the research community to counter the risks. Cybersecurity in Smart Homes is intended for people who wish to understand the architectures, protocols and different technologies used in smart homes.
• Participation de l’équipe TGV à DEFT 2022 : Prédiction automatique de notes d’étudiants à des questionnaires en fonction du type de question
  
  • Gaudray Bouju Vanessa
  • Guettier Margot
  • Lerus Gwennola
  • Guibon Gaël
  • Labeau Matthieu
  • Lefeuvre Luce
  , 2022, pp.23-35. Cet article présente l’approche de l’équipe TGV lors de sa participation à la tâche de base de DEFT 2022, dont l’objectif était de prédire automatiquement les notes obtenues par des étudiants sur la base de leurs réponses à des questionnaires. Notre stratégie s’est focalisée sur la mise au point d’une méthode de classification des questions en fonction du type de réponse qu’elles attendent, de manière à pouvoir mener une approche différenciée pour chaque type. Nos trois runs ont consisté en une approche non différenciée, servant de référence, et deux approches différenciées, la première se basant sur la constitution d’un jeu de caractéristiques et la seconde sur le calcul de TF-IDF et de la fonction de hashage. Notre objectif premier était ainsi de vérifier si des approches dédiées à chaque type de questions sont préférables à une approche globale.
• PBRE: A Rule Extraction Method from Trained Neural Networks Designed for Smart Home Services
  
  • Qiu Mingming
  • Najm Elie
  • Sharrock Rémi
  • Traverson Bruno
  , 2022, 13427, pp.158-173. Designing smart home services is a complex task when multiple services with a large number of sensors and actuators are deployed simultaneously. It may rely on knowledge-based or data-driven approaches. The former can use rule-based methods to design services statically, and the latter can use learning methods to discover inhabitants’ preferences dynamically. However, neither of these approaches is entirely satisfactory because rules cannot cover all possible situations that may change, and learning methods may make decisions that are sometimes incomprehensible to the inhabitant. In this paper, PBRE (Pedagogic Based Rule Extractor) is proposed to extract rules from learning methods to realize dynamic rule generation for smart home systems. The expected advantage is that both the explainability of rule-based methods and the dynamicity of learning methods are adopted. We compare PBRE with an existing rule extraction method, and the results show better performance of PBRE. We also apply PBRE to extract rules from a smart home service represented by an NRL (Neural Network-based Reinforcement Learning). The results show that PBRE can help the NRL-simulated service to make understandable suggestions to the inhabitant. (10.1007/978-3-031-12426-6_13)
  DOI : 10.1007/978-3-031-12426-6_13
• Multiple instance learning on deep features for weakly supervised object detection with extreme domain shifts
  
  • Gonthier Nicolas
  • Ladjal Saïd
  • Gousseau Yann
  Computer Vision and Image Understanding, Elsevier, 2022, 214. Weakly supervised object detection (WSOD) using only image-level annotations has attracted a growing attention over the past few years. Whereas such task is typically addressed with a domain-specific solution focused on natural images, we show that a simple multiple instance approach applied on pre-trained deep features yields excellent performances on non-photographic datasets, possibly including new classes. The approach does not include any fine-tuning or cross-domain learning and is therefore efficient and possibly applicable to arbitrary datasets and classes. We investigate several flavors of the proposed approach, some including multi-layers perceptron and polyhedral classifiers. Despite its simplicity, our method shows competitive results on a range of publicly available datasets, including paintings (People-Art, IconArt), watercolors, cliparts and comics and allows to quickly learn unseen visual categories. (10.1016/j.cviu.2021.103299)
  DOI : 10.1016/j.cviu.2021.103299