Portail HAL de Télécom Paris
Not a member yet
14191 research outputs found
Sort by
New approaches to CLT for stable random variables
No full textInternational audienceIn this paper, we present novel approaches to the Central Limit Theorem (CLT) for stable random variables, particularly focusing on cases where the distribution has heavy tails. The study develops three independent methods, each addressing different aspects of convergence in the context of α-stable distributions. We explore both non-integrable and integrable cases, offering new derivations and highlighting the distinct normalization required when dealing with heavy-tailed distributions. Special attention is given to the Stein method, adapted to handle stable laws, leveraging Fourier techniques and Poisson process representations. We provide convergence rates and quantify the Wasserstein distance between sums of independent and identically distributed random variables and their limiting stable distributions. Our findings extend existing results by offering intrinsic methods for stable CLTs, with applications to distributions such as Pareto, where classical CLT normalization fail
Analytical solution of radiative transfer equation of light radiance in turbid slab with inner-medium source under P3-1D approximation
No full textA generalized model of the 1-dimensional radiative transfer equation of the light radiance in a turbid slad is detailed, under the P-3 approximation, including the possibility to model a continuous plane-wave source located at any depth within the scattering slab. This analytical model, which requires significant evolution of the P3-1D model is extensively described and validated by comparison with Monte-Carlo numerical experiments. A series of numerical simulations illustrates some of the modelling possibilities offered by this extended model, which makes it possible to continuously model the transition between a classical slab geometry and a semi-infinite geometry
Attention‐Based Fusion of Pretrained MaskPoint and Point‐MAE Representations for Enhanced Point Cloud Classification
No full textInternational audienceThis paper proposes a novel attention‐based fusion framework that integrates two self‐supervised learning models, the Point‐MAE and MaskPoint methods, to improve 3D point cloud classification. Each model is pretrained on a carefully selected dataset to optimize downstream performance. Their pretrained encoders are then fused using two attention strategies, namely multi‐layer attention and cross‐attention, to capture complementary features from both networks. A comparative evaluation of their original classification heads revealed that the Point‐MAE architecture, featuring a three‐layer MLP with batch normalization and dropout, consistently outperforms the MaskPoint design and was thus adopted in the proposed framework. Extensive experiments on both synthetic and real‐world benchmarks, including ModelNet40 and multiple variants of ScanObjectNN, demonstrate that attention‐based fusion improves classification robustness in challenging scenarios, with cross‐attention consistently achieving the highest accuracy among the evaluated fusion strategies. Controlled baseline experiments further confirm that these gains stem from meaningful feature interaction. The proposed approach demonstrates that attention‐guided fusion of self‐supervised point cloud representations is an effective and flexible strategy for enhancing downstream 3D recognition performance
Efficient Online Variational Estimation via Monte Carlo Sampling
No full textThis article addresses online variational estimation in parametric state-space models. We propose a new procedure for efficiently computing the evidence lower bound and its gradient in a streaming-data setting, where observations arrive sequentially.The algorithm allows for the simultaneous training of the model parameters and the distribution of the latent states given the observations.It is based on i.i.d. Monte Carlo sampling, coupled with a well-chosen deep architecture, enabling both computational efficiency and flexibility. The performance of the method is illustrated on both synthetic data and real-world air-quality data.The proposed approach is theoretically motivated by the existence of an asymptotic contrast function and the ergodicity of the underlying Markov chain, and applies more generally to the computation of additive expectations under posterior distributions in state-space models
PHYSICS-INFORMED LEARNING OF NEURAL SCATTERING FIELDS TOWARDS MEASUREMENT-FREE MESH-TO-HRTF ESTIMATION
No full textInternational audienceThis paper describes neural simulation of the scattered pressure field from a plane wave around a scattering object in both continuous 2D and 3D domains. This task has typically been treated as a regression problem that aims to train a physicsinformed neural network (PINN) using pressure measurements at discrete positions. This approach, however, needs to train the whole network for each incident wave direction. To address this, we propose a measurement-free simulator based on a PINN purely driven by the Helmholtz equation with the Robin boundary condition and the Sommerfeld radiation condition with the aid of the perfectly matched layer (PML) framework. More specifically, we design a physics-informed scattering hypernetwork (PHISK) that can generalize to incident waves from any direction via low-rank adaptation (LoRA) of a PINN trained for a specific configuration. The experiment shows that the proposed method accurately simulated sound scattering around various objects, adapting to unseen incident wave directions with minimal performance loss, and realized reasonable simulation of head-related transfer functions (HRTFs) from complex mesh data of a human head.Cet article décrit la simulation neuronale du champ de pression diffusée à partir d'une onde plane autour d'un objet diffusant de façon continue à la fois en 2D ou en 3D. Cette tâche a généralement été traitée comme un problème de régression visant à entraîner un réseau de neurones informé par la physique (PINN) à l'aide de mesures acoustiques de la pression à partir d'un nombre fini de positions. Cependant, cette approche nécessite d'entraîner l'ensemble du réseau pour chaque direction d'onde incidente simulée. Pour y remédier, nous proposons un simulateur sans mesure basé sur un PINN purement guidé par l'équation de Helmholtz avec la condition aux limites de Robin et la condition de rayonnement de Sommerfeld, assisté par le cadre de la théorie du 'Perfectly Matched Layer' (PML). Plus précisément, nous concevons un hyperréseau de diffusion informé par la physique nommé PHISK capable de généraliser aux ondes incidentes provenant de n'importe quelle direction via l'adaptation de rang faible (LoRA) d'un PINN entraîné pour une configuration spécifique. L'expérience montre que la méthode proposée simule avec précision la diffusion sonore autour de divers objets, s'adaptant à des directions d'ondes incidentes non vues avec une perte de performance minimale, et réalise une simulation cohérente des fonctions de transfert relatives à la tête (HRTF) à partir de données de maillage complexes d'une tête humaine
Convergence rate for the coupon collector's problem with Stein's method
No full textInternational audienceThe functional characterization of a measure, an essential but delicate aspect of Stein's method, is shown to be accessible for stable probability distributions on convex cones. This notion encompasses the usual stable distributions \textit{e.g.} Gaussian, Pareto, \textit{etc.} but also the max-stable distributions: Weibull, Gumbel and Fréchet. We use the definition of max-stability to define a Markov process whose invariant measure is the stable measure of interest. In this paper, we focus on the Gumbel distribution and show how this construction can be applied to estimate the rate of convergence in the classical coupon collector's problem
Approches d'intégration de structures SIBH et composants de base pour la prochaine génération de transceivers sur phosphure d'indium
No full textData traffic has significantly increased in recent years and is expected to keep rising due to bandwidth-intensive services such as Artificial Intelligence (AI) and high-definition gaming. One approach to enhance transmission data rates is to develop transceivers based on photonic integrated circuits (PICs), which are the key elements responsible for transmitting and receiving signals in Passive Optical Networks (PONs). The fundamental components forming a PIC, known as building blocks (BBs), are the focus of this work.Three approaches were investigated to improve the transmission capabilities of next-generation transceivers. The first concerns the development of SIBH technology used for the fabrication of active BBs. A one-step SIBH structure incorporating a blocking layer was developed. Several studies were conducted to evaluate the influence of growth temperature and layer thickness, with the optimal parameters identified as a growth temperature of 520 °C and a thickness of 40 nm. Additional investigations were also carried out on alternative materials for the semi-insulating layers of the SIBH structure, such as InP:Be and InP:Ru.The second approach focuses on passive BBs. Fabrication processes were developed to enable the monolithic integration of passive and active components despite their two different waveguide architectures (SIBH and deep-ridge). The main challenge encountered was the formation of polycrystals due to the SIBH regrowth, resulting in propagation losses exceeding 10 dB/cm for 1.5 m deep-ridge waveguides. This issue was successfully resolved using an improved second fabrication process.Finally, a new InP (de)multiplexer structure based on a tilted angled multimode interferometer (AMMI) was developed. The device exhibits a compact footprint of 0,09 mm, a crosstalk level of 14.35 dB, and an insertion loss of 8 dB.Le trafic de données a considérablement augmenté ces dernières années et cette tendance devrait se poursuivre avec l'essor de services à forte consommation de bande passante tels que l'Intelligence Artificielle (IA) et les jeux vidéo en haute définition. Une des approches pour accroître les débits de transmission consiste à développer des émetteurs-récepteurs basés sur des circuits intégrés photoniques (PIC), qui constituent les éléments de transmission et de réception dans les réseaux optiques passifs (PON). Les composants élémentaires formant un PIC, appelés briques de base (BB), sont au cœur de ce travail.Trois axes de recherche ont été explorés pour améliorer les capacités de transmission des émetteurs-récepteurs. Le premier concerne le développement de la technologie SIBH utilisée pour la fabrication des BB actifs. Une structure SIBH en une seule étape, intégrant une couche de blocage, a été développée. Plusieurs études ont été menées sur la température de croissance et l'épaisseur de cette couche afin d'identifier les paramètres optimisant les performances des composants. Les meilleures conditions obtenues sont une température de croissance de 520°C et une épaisseur de 40 nm. En complément, d'autres matériaux destinés aux couches semi-isolantes de la structure SIBH, tels que InP:Be et InP:Ru, ont été investigués.Le deuxième axe concerne les BB passifs. Des procédés de fabrication ont été étudiés afin de permettre l'intégration monolithique de composants passifs et actifs malgré l'utilisation de deux architectures de guides d'onde différentes (SIBH et deep-ridge). Le principal obstacle rencontré a été la présence de polycristaux liés à la recroissance SIBH, entraînant une augmentation des pertes de propagation de plus de 10 dB/cm pour des guides deep-ridge de 1,5 µm. Ce problème a pu être résolu grâce à un second procédé de fabrication optimisé.Enfin, une nouvelle structure de (dé)multiplexeur en InP basée sur un interféromètre multimode incliné (AMMI) a été développée. Le dispositif obtenu présente une empreinte réduite de 0,09 mm2, une diaphonie de 14,35 dB et une perte d'insertion de 8 dB
A POP ⋆ is Born: Formal Predictable Out-of-Order Processor Model
No full textModern processors, even at the mid-range level, include multi-level caches, pipelines with branch predictors, or Out-of-Order (OoO) execution. While these are essential for average-case performance, they also increase the complexity of worst-case execution time analysis. OoO execution, for instance, is prone to timing anomalies and, due to the lack of efficient abstractions, quickly leads to state-space explosion. Consequently, it remains highly challenging in the context of critical real-time systems.This work proposes the first generic approach to predictable OoO execution, which is formally modeled and proven in the F* language and experimentally evaluated through simulations in gem5. Performance is evaluated on different processor models for MiBench and Embench programs. The average slowdown for an ARM A710-like processor model amounts to about 18.3% due to an implementation particularity of gem5. Eliminating bias from this issue reduces the slowdown to only 8.8%-10.4%
Group Conversational Agents: A Review of Designs that Support and Shape Group Interaction
No full textInternational audienceConversational agents that participate in or mediate group interaction introduce challenges that extend beyond supporting individual users, raising new questions about how agents participate in and influence groups. To characterise this emerging design space, we present a systematic review of 53 peer-reviewed studies on group conversational agents (GCAs). We analyse how GCAs intervene in group-level processes, including participation regulation, conflict mediation, task alignment, and execution support. Using concepts from group research as an analytic lens, we organise prior GCA work around recurring group interactional challenges (orientation, conflict, alignment, and execution), and examine the roles agents are designed to play in addressing these challenges. We find that GCAs are predominantly designed as short-term, role-bounded interventions targeting isolated challenges in bounded interactional contexts. We further identify recurring structural tensions in GCA design, including tradeoffs between visibility and discretion, proactivity and group autonomy, and agent authority and group ownership. Together, these findings clarify how current GCAs are positioned within group interaction, surface the implicit assumptions embedded in their designs, and outline open questions for future research on conversational agents as group-level interventions
UNSUPERVISED DOMAIN ADAPTATION WITH TARGET-ONLY MARGIN DISPARITY DISCREPANCY
No full textInternational audienceIn interventional radiology, Cone-Beam Computed Tomography (CBCT) is a helpful imaging modality that provides guidance to practicians during minimally invasive procedures. CBCT differs from traditional Computed Tomography (CT) due to its limited reconstructed field of view, specific artefacts, and the intra-arterial administration of contrast medium. While CT benefits from abundant publicly available annotated datasets, interventional CBCT data remain scarce and largely unannotated, with existing datasets focused primarily on radiotherapy applications. To address this limitation, we leverage a proprietary collection of unannotated interventional CBCT scans in conjunction with annotated CT data, employing domain adaptation techniques to bridge the modality gap and enhance liver segmentation performance on CBCT. We propose a novel unsupervised domain adaptation (UDA) framework based on the formalism of Margin Disparity Discrepancy (MDD), which improves target domain performance through a reformulation of the original MDD optimization framework. Experimental results on CT and CBCT datasets for liver segmentation demonstrate that our method achieves state-of-the-art performance in UDA, as well as in the few-shot setting.</div