Publications

2023

• Noise, Dynamics and Squeezed Light in Quantum Dot and Interband Cascade Lasers
  
  • Zhao Shiyuan
  , 2023. Semiconductor lasers have become ubiquitous in both scientific research and engineering applications, and their miniaturization has made significant strides since their initial demonstration in 1960. Two prominent advancements in this domain include quantum dot (QD) lasers, which operate in the near-infrared wavelength range, and interband cascade lasers (ICLs), designed for mid-infrared operation. Two prominent advancements in this domain include quantum dot (QD) lasers, which operate in the near-infrared wavelength range, and interband cascade lasers (ICLs), designed for mid-infrared operation. In the current landscape of optoelectronics, photonic integrated circuits (PICs) play a pivotal and far-reaching role. They offer unmatched scalability, reduced weight, cost-effectiveness, and energy efficiency by enabling the fabrication of complete optical systems using versatile building blocks seamlessly integrated onto a single chip. In this context, the direct epitaxial growth of III-V materials on silicon holds promise as a compelling approach for the development of coherent laser sources. QD lasers with their ultimate three-dimensional carrier confinement, high thermal stability, and robust tolerance for epitaxial defects are promising candidates for serving as on-chip laser sources. Additionally, ICLs are also well-suited for integration into silicon, making them ideal for compact chemical sensing systems. Noise considerations are indeed paramount when it comes to assessing the quality and reliability of technologies. Achieving the shot noise limit and the Schawlow-Townes linewidth has long been recognized as significant milestones. To tackle noise issues, a range of noise reduction techniques has been explored, encompassing passive optical feedback within an external cavity and active electronic feedback mechanisms to compensate for injection current fluctuations. However, while feedback systems can mitigate laser noise, they can also introduce more intricate nonlinear dynamics, giving rise to phenomena like periodic oscillation, square-wave oscillation, and chaos. The first part of this thesis involves an in-depth investigation into noise and dynamics in two distinct laser types. QD lasers are found to exhibit a high degree of robustness when exposed to parasitic optical reflections but manifest increased sensitivity to optoelectronic feedback. Conversely, ICLs display a spectrum of dynamic behaviours when subjected to optical feedback. Furthermore, recent advancements in low-noise pumping circuits for lasers have led to the generation of amplitude-squeezed light. This represents a transition from classical noise to quantum noise, opening up new possibilities in the field of laser technology and quantum optics. The second part of this thesis delves into the phenomenon of amplitude squeezing in both QD lasers and ICLs. The findings indicate that both types of lasers can exhibit broadband squeezing bandwidth and a significant level of squeezing. All these outcomes in this study contribute to a deeper comprehension of the characteristics of QD lasers and ICLs, laying the groundwork for the development of high-performance classical and quantum emitters on PICs in the future.
• The Locality and Symmetry of Positional Encodings
  
  • Chen Lihu
  • Varoquaux Gaël
  • Suchanek Fabian M.
  , 2023. Positional Encodings (PEs) are used to inject word-order information into transformer-based language models. While they can significantly enhance the quality of sentence representations, their specific contribution to language models is not fully understood, especially given recent findings that various positional encodings are insensitive to word order. In this work, we conduct a systematic study of positional encodings in Bidirectional Masked Language Models (BERT-style) , which complements existing work in three aspects: (1) We uncover the core function of PEs by identifying two common properties, Locality and Symmetry; (2) We show that the two properties are closely correlated with the performances of downstream tasks; (3) We quantify the weakness of current PEs by introducing two new probing tasks, on which current PEs perform poorly. We believe that these results are the basis for developing better PEs for transformer-based language models. The code is available at https://github.
• Proceedings of the Workshop on Deep Learning for Knowledge Graphs (DL4KG 2023) co-located with the 21st International Semantic Web Conference (ISWC 2023)
  
  • Alam Mehwish
  • Buscaldi Davide
  • Cochez Michael
  • Osborne Francesco
  • Recupero Diego Reforgiato
  , 2023, 3559.
• Fair Text Classification with Wasserstein Independence
  
  • Leteno Thibaud
  • Gourru Antoine
  • Laclau Charlotte
  • Emonet Rémi
  • Gravier Christophe
  , 2023, pp.15790-15803. Group fairness is a central research topic in text classification, where reaching fair treatment between sensitive groups (e.g. women vs. men) remains an open challenge. This paper presents a novel method for mitigating biases in neural text classification, agnostic to the model architecture. Considering the difficulty to distinguish fair from unfair information in a text encoder, we take inspiration from adversarial training to induce Wasserstein independence between representations learned to predict our target label and the ones learned to predict some sensitive attribute. Our approach provides two significant advantages. Firstly, it does not require annotations of sensitive attributes in both testing and training data. This is more suitable for real-life scenarios compared to existing methods that require annotations of sensitive attributes at train time. Secondly, our approach exhibits a comparable or better fairness-accuracy trade-off compared to existing methods. (10.18653/v1/2023.emnlp-main.978)
  DOI : 10.18653/v1/2023.emnlp-main.978
• Connecting Symbolic Statutory Reasoning with Legal Information Extraction
  
  • Holzenberger Nils
  • van Durme Benjamin
  , 2023, pp.113-131. Statutory reasoning is the task of determining whether a given law-a part of a statute-applies to a given legal case. Previous work has shown that structured, logical representations of laws and cases can be leveraged to solve statutory reasoning, including on the StAtutory Reasoning Assessment dataset (SARA), but rely on costly human translation into structured representations. Here, we investigate a form of legal information extraction atop the SARA cases, illustrating how the task can be done with high performance. Further, we show how the performance of downstream symbolic reasoning directly correlates with the quality of the information extraction. (10.18653/v1/2023.nllp-1.12)
  DOI : 10.18653/v1/2023.nllp-1.12
• Automatic Analysis of Substantiation in Scientific Peer Reviews
  
  • Guo Yanzhu
  • Shang Guokan
  • Rennard Virgile
  • Vazirgiannis Michalis
  • Clavel Chloé
  , 2023, pp.10198-10216. With the increasing amount of problematic peer reviews in top AI conferences, the community is urgently in need of automatic quality control measures. In this paper, we restrict our attention to substantiation — one popular quality aspect indicating whether the claims in a review are sufficiently supported by evidence — and provide a solution automatizing this evaluation process. To achieve this goal, we first formulate the problem as claim-evidence pair extraction in scientific peer reviews, and collect SubstanReview, the first annotated dataset for this task. SubstanReview consists of 550 reviews from NLP conferences annotated by domain experts. On the basis of this dataset, we train an argument mining system to automatically analyze the level of substantiation in peer reviews. We also perform data analysis on the SubstanReview dataset to obtain meaningful insights on peer reviewing quality in NLP conferences over recent years. The dataset is available at https://github.com/YanzhuGuo/SubstanReview. (10.18653/v1/2023.findings-emnlp.684)
  DOI : 10.18653/v1/2023.findings-emnlp.684
• Toward Stronger Textual Attack Detectors
  
  • Colombo Pierre
  • Picot Marine
  • Noiry Nathan
  • Staerman Guillaume
  • Piantanida Pablo
  , 2023, pp.484-505. The landscape of available textual adversarial attacks keeps growing, posing severe threats and raising concerns regarding the deep NLP system's integrity. However, the crucial problem of defending against malicious attacks has only drawn the attention of the NLP community. The latter is nonetheless instrumental in developing robust and trustworthy systems. This paper makes two important contributions in this line of search: (i) we introduce LAROUSSE, a new framework to detect textual adversarial attacks and (ii) we introduce STAKEOUT, a new benchmark composed of nine popular attack methods, three datasets, and two pre-trained models. LAROUSSE is ready-to-use in production as it is unsupervised, hyperparameter-free, and non-differentiable, protecting it against gradient-based methods. Our new benchmark STAKEOUT allows for a robust evaluation framework: we conduct extensive numerical experiments which demonstrate that LAROUSSE outperforms previous methods, and which allows to identify interesting factors of detection rate variations. (10.18653/v1/2023.findings-emnlp.35)
  DOI : 10.18653/v1/2023.findings-emnlp.35
• Enhancements in Embedded Systems Security using Machine Learning
  
  • Shrivastwa Ritu Ranjan
  , 2023. The list of connected devices (or IoT) is growing longer with time and so is the intense vulnerability to security of the devices against targeted attacks originating from network or physical penetration, popularly known as Cyber Physical Security (CPS) attacks. While security sensors and obfuscation techniques exist to counteract and enhance security, it is possible to fool these classical security countermeasures with sophisticated attack equipment and methodologies as shown in recent literature. Additionally, end node embedded systems design is bound by area and is required to be scalable, thus, making it difficult to adjoin complex sensing mechanism against cyberphysical attacks. The solution may lie in Artificial Intelligence (AI) security core (soft or hard) to monitor data behaviour internally from various components. Additionally the AI core can monitor the overall device behaviour, including attached sensors, to detect any outlier activity and provide a smart sensing approach to attacks. AI in hardware security domain is still not widely acceptable due to the probabilistic behaviour of the advanced deep learning techniques, there have been works showing practical implementations for the same. This work is targeted to establish a proof of concept and build trust of AI in security by detailed analysis of different Machine Learning (ML) techniques and their use cases in hardware security followed by a series of case studies to provide practical framework and guidelines to use AI in various embedded security fronts. Applications can be in PUFpredictability assessment, sensor fusion, Side Channel Attacks (SCA), Hardware Trojan detection, Control flow integrity, Adversarial AI, etc.
• Verifiable Decentralized Multi-Client Functional Encryption for Inner Product
  
  • Nguyen Duy
  • Phan Duong Hieu
  • Pointcheval David
  , 2023. Joint computation on encrypted data is becoming increasingly crucial with the rise of cloud computing. In recent years, the development of multi-client functional encryption (MCFE) has made it possible to perform joint computation on private inputs, without any interaction. Well-settled solutions for linear functions have become efficient and secure, but there is still a shortcoming: if one user inputs incorrect data, the output of the function might become meaningless for all other users (while still useful for the malicious user). To address this issue, the concept of verifiable functional encryption was introduced by Badrinarayanan et al. at Asiacrypt '16 (BGJS). However, their solution was impractical because of strong statistical requirements. More recently, Bell et al. introduced a related concept for secure aggregation, with their ACORN solution, but it requires multiple rounds of interactions between users. In this paper, we first propose a computational definition of verifiability for MCFE. Our notion covers the computational version of BGJS and extends it to handle any valid inputs defined by predicates. The BGJS notion corresponds to the particular case of a fixed predicate in our setting; we then introduce a new technique called Combine-then-Descend, which relies on the class group. It allows us to construct One-time Decentralized Sum (ODSUM) on verifiable private inputs. ODSUM is the building block for our final protocol of a verifiable decentralized MCFE for inner-product, where the inputs are within a range. Our approach notably enables the efficient identification of malicious users, thereby addressing an unsolved problem in ACORN.
• Algebraic Attacks on Round-Reduced RAIN and Full AIM-III
  
  • Zhang Kaiyi
  • Wang Qingju
  • Yu Yu
  • Guo Chun
  • Cui Hongrui
  , 2023.
• Unified Measures for the Rate-Distortion-Latency Trade-Off
  
  • Vijayaratnam Melan
  • Milovanovic Marta
  • Cagnazzo Marco
  • Tartaglione Enzo
  • Valenzise Giuseppe
  , 2023. In today's digital age, multimedia content is omnipresent, and the demand for efficient compression techniques is everincreasing. In particular, the successful delivery of services based on video transmission largely depends on achieving the lowest latency values. One solution has been to use extrapolation for latency compensation in video transmission that allows to reduce the latency by an arbitrary amount. Nevertheless, this latency reduction comes at the cost of an increased distortion of the displayed images, since they are based on temporal extrapolation. Latency can also be traded with coding rate. This paper introduces ELR-PSNR and EPR-Latency as unified metrics to assess the three-way trade-off between rate, distortion, and latency simultaneously.
• Comparing a Mentalist and an Interactionist Approach for Trust Analysis in Human-Robot Interaction
  
  • Hulcelle Marc
  • Varni Giovanna
  • Rollet Nicolas
  • Clavel Chloé
  , 2023, pp.273-280. (10.1145/3623809.3623840)
  DOI : 10.1145/3623809.3623840
• DS-IRSA: A Deep Reinforcement Learning and Sensing Based IRSA
  
  • Hmedoush Iman
  • Gu Pengwenlong
  • Adjih Cédric
  • Mühlethaler Paul
  • Serhrouchni Ahmed
  , 2023. One of the main difficulties to enable the future scaling of IoT networks is the issue of massive connectivity. Recently, Modern Random Access protocols have emerged as a promising solution to provide massive connections for IoT. One main protocol of this family is Irregular Repetition Slotted Aloha (IRSA), which can asymptotically reach the optimal throughput of 1 packet/slot. Despite this, the problem is not yet solved due to lower throughput in non-asymptotic cases with smaller frame sizes. In this paper, we propose a new variant of IRSA protocol named Deep-Learning and Sensing-based IRSA (DS-IRSA) to optimise the performance of IRSA in short frame IoTs, where a sensing phase is added before the transmission phase and users' actions in both phases are managed by a deep reinforcement learning (DRL) method. Our goal is to learn to interact and ultimately to learn a sensing protocol entirely through Deep Learning. In this way, active users can coordinate well with each other and the throughput of the whole system can be well improved. Simulation results show that our proposed scheme convergence quickly towards the optimal performance of almost 1 packet/slot for small frame sizes and with enough minislots and can achieve higher throughput in almost all cases.
• A closed-measure approach to stochastic approximation
  
  • Bianchi Pascal
  • Rios-Zertuche Rodolfo
  Stochastics: An International Journal of Probability and Stochastic Processes, Taylor & Francis: STM, Behavioural Science and Public Health Titles, 2023, pp.1-23. This paper introduces a new method to tackle the issue of the almost sure convergence of stochastic approximation algorithms defined from a differential inclusion. Under the assumption of slowly decaying step-sizes, we establish that the set of essential accumulation points of the iterates belongs to the Birkhoff center associated with the differential inclusion. Unlike previous works, our results do not rely on the notion of asymptotic pseudotrajectories introduced by Benaı̈m–Hofbauer–Sorin, which is the predominant technique to address the convergence problem. They follow as a consequence of Young’s superposition principle for closed measures. This perspective bridges the gap between Young’s principle and the notion of invariant measure of set-valued dynamical systems introduced by Faure and Roth. Also, the proposed method allows to obtain sufficient conditions under which the velocities locally compensate around any essential accumulation point. (10.1080/17442508.2024.2353278)
  DOI : 10.1080/17442508.2024.2353278
• 2D indoor localization resolution improvement using IR-UWB stepped frequencies signals
  
  • Cousin Jean-Christophe
  • Muller Muriel
  • Awarkeh Nour
  • Samama Nel
  IEEE Sensors Letters, IEEE, 2023, 7 (12), pp.3503404. This article discusses the ability to improve the resolution of range and azimuth angle measurements by an interferometric localization system used to locate beacons within its line of sight (LoS) by exploiting stepped frequency signals Impulse Radio - Ultra WideBand (IR-UWB) centered on 2 frequencies. This improvement does not require increasing the frequency bandwidth used. This solution uses a phase correlation (PC) method, usually applied to continuous wave (CW) signals, adapted to Ultra WideBand (UWB) pulse signals. The results obtained are compared with those computed by a classical method of energy detection (ED) by exploiting the same frequency band. (10.1109/LSENS.2023.3328793)
  DOI : 10.1109/LSENS.2023.3328793
• Le facteur est bien passé
  
  • Zayana Karim
  • Braun Nathalie
  CultureMath, ENS, 2023. Développer ou factoriser une expression algébrique est une compétence de base en mathématiques. Les programmes du premier degré y préparent tandis que ceux du second la renforcent et l'automatisent. Nous récapitulons dans cet article les formules à connaître et donnons à les visualiser grâce à quelques images.
• Some power allocation algorithms for cognitive uplink satellite systems
  
  • Louchart Arthur
  • Tohidi Ehsan
  • Gesbert David
  • Ciblat Philippe
  • Lagunas Eva
  • Poulliat Charly
  EURASIP Journal on Wireless Communications and Networking, SpringerOpen, 2023, 2023: 32, pp.1-30. Cognitive satellite communication (SatCom) is rapidly emerging as a promising technology to overcome the scarcity of the exclusive licensed band model in order to fulfill the increasing demand for high data rate services. The paper addresses power allocation methods for multi-operator multi-beam uplink satellite communication systems co-existing with a Ka-band terrestrial network, using cognitive radio paradigm. Such a scenario is especially challenging because of (i) the coexisting multiple SatCom operators over the cognitive band need to coordinate the use of their resources under limited inter-operator information exchange, and (ii) nonlinear onboard high power amplifier (HPA) which leads to nonlinear interference between users and beams. In order to tackle the first challenge, we propose distributed power allocation algorithms including the standard Alternate Direction Multiplier Method (ADMM); Regarding the HPA nonlinear impairment, we propose nonlinear-aware power allocation based on Signomial Programming. The proposed solutions outperform state-of-the-art in both cases. (10.1186/s13638-023-02234-7)
  DOI : 10.1186/s13638-023-02234-7
• Le Soleil se lèvera-t-il demain ?
  
  • Zayana Karim
  • Rioul Olivier
  Tangente (Paris), 2023, pp.46-47. Faut-il vivre chaque jour comme s’il était le dernier ? Cette question est légitime quand on sait que Pierre-Simon de Laplace, en 1814, évaluait à 0.99999945 la probabilité que le Soleil se lève à nouveau le lendemain. Dans cet article, nous expliquons de façon très détaillée quels sont les raisonnements et les calculs qui conduisent à ce résultat pour le moins surprenant.
• Resilient Multiple Choice Learning: A learned scoring scheme with application to audio scene analysis
  
  • Letzelter Victor
  • Fontaine Mathieu
  • Chen Mickaël
  • Pérez Patrick
  • Essid Slim
  • Richard Gael
  , 2023. We introduce Resilient Multiple Choice Learning (rMCL), an extension of the MCL approach for conditional distribution estimation in regression settings where multiple targets may be sampled for each training input. Multiple Choice Learning is a simple framework to tackle multimodal density estimation, using the Winner-Takes-All (WTA) loss for a set of hypotheses. In regression settings, the existing MCL variants focus on merging the hypotheses, thereby eventually sacrificing the diversity of the predictions. In contrast, our method relies on a novel learned scoring scheme underpinned by a mathematical framework based on Voronoi tessellations of the output space, from which we can derive a probabilistic interpretation. After empirically validating rMCL with experiments on synthetic data, we further assess its merits on the sound source localization problem, demonstrating its practical usefulness and the relevance of its interpretation.
• Trois lois des erreurs
  
  • Rioul Olivier
  Tangente (Paris), 2023, pp.48-50. Trois lois des erreurs de Laplace L'observation des astres est toujours soumise à des erreurs. Comment choisir la valeur « la plus pertinente » parmi plusieurs mesures ? Laplace, mais aussi Legendre et Gauss, échafaudent des théories qui conduisent in fine à la fameuse loi normale.
• A Compact and Semantic Latent Space for Disentangled and Controllable Image Editing
  
  • Lesné Gwilherm
  • Gousseau Yann
  • Ladjal Saïd
  • Newson Alasdair
  Proceedings of the 20th ACM SIGGRAPH European Conference on Visual Media Production, 2023, pp.1-10. Recent advances in the field of generative models and in particular generative adversarial networks (GANs) have lead to substantial progress for controlled image editing, especially compared with the pre-deep learning era. Despite their powerful ability to apply realistic modifications to an image, these methods often lack properties like disentanglement (the capacity to edit attributes independently). In this paper, we propose an auto-encoder which reorganizes the latent space of StyleGAN, so that each attribute which we wish to edit corresponds to an axis of the new latent space, and furthermore that the latent axes are decorrelated, encouraging disentanglement. We work in a compressed version of the latent space, using Principal Component Analysis, meaning that the parameter complexity of our autoencoder is reduced, leading to short training times (∼ 45 mins). Qualitative and quantitative results demonstrate the editing capabilities of our approach, with greater disentanglement than competing methods, while maintaining fidelity to the original image with respect to identity. Our autoencoder architecture simple and straightforward, facilitating implementation. (10.1145/3626495.3626508)
  DOI : 10.1145/3626495.3626508
• Statistical Understanding of Adversarial Robustness
  
  • Goibert Morgane
  , 2023. This thesis focuses on the question of robustness in machine learning, specifically examining two types of attacks: poisoning attacks at training time and evasion attacks at inference time.The study of poisoning attacks dates back to the sixties and has been unified under the theory of robust statistics. However, prior research was primarily focused on classical data types, mainly real-numbered data, limiting the applicability of poisoning attack studies. In this thesis, robust statistics are extended to ranking data, which lack a vector space structure and have a combinatorial nature. The work presented in this thesis initiates the study of robustness in the context of ranking data and provides a framework for future extensions. Contributions include a practical algorithm to measure the robustness of statistics for the task of consensus ranking, and two robust statistics to solve this task.In contrast, since 2013, evasion attacks gained significant attention in the deep learning field, particularly for image classification. Despite the proliferation of research works on adversarial examples, the theoretical analysis of the problem remains challenging and it lacks unification. To address this matter, the thesis makes contributions to understanding and mitigating evasion attacks. These contributions involve the unification of adversarial examples' characteristics through the study of under-optimized edges and information flow within neural networks, and the establishment of theoretical bounds characterizing the success rate of modern low-dimensional attacks for a wide range of models.
• Traitement de la phase des signaux audio dans les réseaux de neurones profonds
  
  • Mathieu Félix
  , 2023. La tâche de séparation de sources sonores d'un enregistrement audio requiert un traitement tout particulier. L'avènement des réseaux de neurones profonds a permis d'améliorer cette tâche au prix d'une complexité computationnelle accrue et d'une opacité des algorithmes. Les interférences induites par ces algorithmes, qu'elles soient parasites ou structurées, peuvent perturber la compréhension du signal, en particulier dans le contexte de la restitution de la voix. Ces problèmes se manifestent particulièrement lors de la transmission de discussions en temps réel, exigeant des mesures de performance pour évaluer les modèles de séparation de sources. Les critères incluent la qualité de reconstruction des pistes individuelles, l'intelligibilité des signaux vocaux, la résilience face aux interférences, et d'autres aspects tels que la réduction des coûts computationnels et l'interprétabilité des traitements. Cette thèse vise à rendre ces modèles plus interprétables tout en atténuant leur coût computationnel, en se concentrant particulièrement sur la modélisation de la phase des signaux. La difficulté actuelle réside dans la modélisation adéquate de cette composante, cruciale pour la compréhension du signal audio. Nous explorerons des stratégies telles que l'utilisation de modèles à valeurs complexes, de représentations invariantes à la phase, et de modèles permettant de s'abstraire de la composante de phase. L'objectif final est de parvenir à des avancées significatives dans la modélisation de la phase des signaux au sein des réseaux de neurones profonds, tout en préservant ou réduisant les coûts computationnels et en améliorant l'interprétabilité des décisions des algorithmes existants.
• CryptoConcurrency: (Almost) Consensusless Asset Transfer with Shared Accounts
  
  • Tonkikh Andrei
  • Ponomarev Pavel
  • Kuznetsov Petr
  • Pignolet Yvonne-Anne
  , 2023, pp.1556-1570. A typical blockchain protocol uses consensus to make sure that mutually mistrusting users agree on the order in which their operations on shared data are executed. However, it is known that asset transfer systems, by far the most popular application of blockchains, can be implemented without consensus. Assuming that no account can be accessed concurrently and every account belongs to a single owner, one can efficiently implement an asset transfer system in a purely asynchronous, consensus-free manner. It has also been shown that implementing asset transfer with shared accounts is impossible without consensus. In this paper, we propose CryptoConcurrency, an asset transfer protocol that allows concurrent accesses to be processed in parallel, without involving consensus, whenever possible. More precisely, if concurrent transfer operations on a given account do not lead to overspending, i.e. can all be applied without the account balance going below zero, they proceed in parallel. Otherwise, the account's owners may have to access an external consensus object. Notably, we avoid relying on a central, universally-trusted, consensus mechanism and allow each account to use its own consensus implementation, which only the owners of this account trust. This provides greater decentralization and flexibility. (10.1145/3576915.3616587)
  DOI : 10.1145/3576915.3616587
• Convergence and robustness of the Hopf oscillator applied to an ABLE exoskeleton: reachability analysis and experimentation
  
  • Hafs Abdelwaheb
  • Verdel Dorian
  • Jerray Jawher
  • Bruneau Olivier
  • Vignais Nicolas
  • Berret Bastien
  • Fribourg Laurent
  , 2023. Convergence and robustness of the Hopf oscillator applied to an ABLE exoskeleton: reachability analysis and experimentation