HAL Portal UPHF (Université Polytechnique Hauts-de-France)
Not a member yet
28473 research outputs found
Sort by
Noise sources, data process and uncertainties in terahertz time domain spectroscopy
No full textInternational audienceThe first terahertz time domain spectroscopy (THz-TDS) experiments are now well-established, and fully autonomous systems have been commercially available for over a decade. This has led to the increasing adoption of THz-TDS by research groups outside the core THz community, as well as by industrial actors. With this broader use, the need for more standardized experimental and data processing methods has become crucial, exemplified by the publication of standards for specifying THz-TDS systems. THz-TDS uniquely performs measurements in the time domain, while the information of interest often resides in the frequency domain. This necessitates a minimum level of data processing, typically starting with a fast Fourier transform. Various strategies exist to extract the desired information, and the field continues to evolve rapidly, lacking a definitive standard. In this conference, we will briefly discuss the different sources of noise and uncertainties in a THz-TDS setup, followed by a review of proposed methodologies for extracting information from time traces, and finally, a discussion on obtaining better insight into the uncertainties of the retrieved information
Effects of La2O3 on structure and mechanical properties of sodium borosilicate glass
No full textInternational audienceTo develop rare-earth-doped borosilicate glasses with enhanced mechanical properties, lanthanum sodium borosilicate glasses were successfully synthesized using a fusion method. The glass composition was (mol.%): 55SiO2 - 25B2O3 – (20-x)Na2O - xLa2O3 (x = 1, 2, 3, 4, and 5). The primary aim was to investigate the impact of Na2O substitution by La2O3 on the structural and mechanical characteristics of the glasses. X-ray diffraction analysis confirmed the amorphous nature of the samples, while Raman and Fourier-transform infrared spectroscopy indicated a conversion of BO3 to BO4 units. Additionally, differential thermal analysis revealed an increase in both glass transition and crystallization temperatures with higher La2O3 content. The inclusion of La2O3 enhanced the glass density, Vickers hardness, fracture toughness, and Young's modulus, owing to the larger ionic radius of La3+ compared to other components. In summary, the incorporation of an optimal concentration of La2O3 significantly improves the mechanical properties of these sodium borosilicate glasses, making them more suitable for applications involving the containment of actinides in nuclear fuel reprocessing
SnatchML: Hijacking ML Models Without Training Access
No full textOur code is available at https://github.com/ihsenLab/SnatchML.International audienceThe widespread deployment of Machine Learning (ML) models has been accompanied by the emergence of various attacks that threaten their trustworthiness and raise ethical and societal concerns. One such attack is model hijacking, where an adversary seeks to repurpose a victim model to perform a different task than originally intended. Model hijacking can cause significant accountability and security risks since the owner of a hijacked model can be framed for having their model offer illegal or unethical services. Prior works consider model hijacking as a training time attack, whereby an adversary requires full access to the ML model training. In this paper, we consider a stronger threat model for an inference-time hijacking attack, where the adversary has no access to the training phase of the victim model. Our intuition is that ML models, which are typically over-parameterized, might have the capacity to (unintentionally) learn more than the intended task they are trained for. We propose SnatchML, a new training-free model hijacking attack, that leverages the extra capacity learnt by the victim model to infer different tasks that can be semantically related or unrelated to the original one. Our results on models deployed on AWS Sagemaker showed that SnatchML can deliver high accuracy on hijacking tasks. Interestingly, while all previous approaches are limited by the number of classes in the benign task, SnatchML can hijack models for tasks that contain more classes than the original. We explore different methods to mitigate this risk; We propose meta-unlearning, which is designed to help the model unlearn a potentially malicious task while training for the original task. We also provide insights on over-parametrization as a possible inherent factor that facilitates model hijacking, and accordingly, we propose a compression-based countermeasure to counteract this attack. We believe this work offers a previously overlooked perspective on model hijacking attacks, presenting a stronger threat model and higher applicability in real-world contexts. Our code is available at https://github.com/ihsenLab/SnatchML
Marceline Desbordes-Valmore : canadienne sans le savoir ?: Étude sur la réception de Marceline Desbordes-Valmore par le Canada francophone de la fin du XIXe siècle au début du XXe siècle ?
No full textInternational audienceThe influence of Marceline Desbordes-Valmore's poetry on modernity is significant, though discreet. An author who achieved dazzling success during her lifetime, further enhanced by a critically acclaimed posthumous collection, the poet of Les Pleurs was unable to sustain her relative success.This writer had a definite, but little-documented, influence on Quebec poetry. Indeed, in several of his critical articles on Quebec literature, Louis Dantin, known for having edited the works of Émile Nelligan, mentions her several times. First, to describe a poem by Alice Lemieux-Levesque entitled "To Marcelline [sic] Desbordes-Valmore," which he finds rather pallidHe also mentions her in one of his articles on Blanche Lamontagne. He says that this other Quebec scholar does not exude the great passion "of Louisa Siefert or the restless Marcelline [sic]," making her the one who hums "the honest and calm love of the good people from Gaspésie."During the annual conference organised by the Association for Canadian and Québec Literatures ( ACQL), I had the privilege to talk about the presence of Marceline Desbordes-Valmore's works in French-canadian poetry from the end of the 19th century, to the beginning of the 20th century. I was able to link Marceline Desbordes-Valmore's works with those from French-Canadian authors such as Louis Dantin, Robert Choquette, Alice Lemieux-Levesque, Blanche Lamontagne, and Émile Nelligan, to name a few.Similarly, even if we know the influence that Rimbaud and Baudelaire had on Émile Nelligan, questioning the potential influence of the poet who inspired these two authors on the Canadian poet seems convincing to us.L’influence de la poésie de Marceline Desbordes-Valmore sur la modernité est non négligeable, bien que discrète. Parmi tous les grands noms du XIXe siècle, cette femme poétesse n’apparaît, de prime abord, pas comme un auteur de premier plan. Peu visible dans la critique dix-neuviémiste, sauf dans certains contextes précis, l’œuvre de Marceline Desbordes-Valmore a bénéficié, à tort ou à raison, d’une réception contrastée.Auteur au succès fulgurant de son vivant, redoublé par un recueil posthume bien accueilli par la critique, la poétesse des Pleurs n’a pas su pérenniser son succès tout relatif.Cette femme de lettres a eu une influence certaine, mais peu documentée, sur la poésie québécoise. En effet, dans plusieurs de ses articles critiques de la littérature québécoise, Louis Dantin, connu pour avoir édité les œuvres d’Émile Nelligan, en fait mention plusieurs fois.Premièrement, pour décrire un poème d’Alice Lemieux-Levesque intitulé « À Marcelline [sic] Desbordes-Valmore », qu’il trouve plutôt pâle. Ce dernier fait aussi, dans un de ses articles consacrés à Blanche Lamontagne, une comparaison avec la poétesse française. Ainsi dit-il que cetteautre lettrée québécoise n’exhale pas la grande passion « de Louisa Siefert ou de l’inquiète Marcelline [sic] », faisant d’elle celle qui fredonne « l’amour honnête et calme des braves gens de la Gaspésie ».Lors de l'édition 2025 du colloque annuel organisé par l'Association des Littératures Canadienne et Québécoise (ALCQ), nous avons eu l'honneur d'aborder, à travers l'analyse de plusieurs poètes et poétesses canadiens (Blanche Lamontagne, Alice Lemieux-Lévesque, Louis Dantin, Robert Choquette, et Émile Nelligan, pour ne citer qu'eux), l’influence qu’a eu la poésie desbordes-valmorienne dans le Canada francophone du début du XXe siècle
Topology optimization of phoxonic crystals for maximizing dual bandgaps using GA-SIMP method
No full textInternational audiencePhoxonic crystals are photonic-phononic or optomechanical periodic structures that simultaneously exhibit dual bandgaps. This allows for the confinement of both optical and elastic waves in cavities and waveguides, providing a powerful platform for novel optomechanical devices and systems. The opening of dual bandgaps is crucial to these potential applications. Topology optimization offers maximum freedom in the dual bandgap structure design, however, relevant research is limited. We propose a two-stage algorithm to maximize the dual bandgaps, where a genetic algorithm is used to find the initial design, and then the SIMP method is employed to obtain the optimal solution. This GA-SIMP hybrid approach capitalizes on the global search capability of GA to explore the design space and identify potential configurations, while harnessing the computational efficiency and precision of SIMP to refine and converge to high-quality solutions. This strategy effectively balances global exploration with local refinement, addressing the trade-off challenges in dual bandgap optimization for phoxonic crystals. We demonstrate the design capability of the coupled methodology by opening bandgaps between different bands in the numerical examples, and the optimized structures show intermediate states between interconnected and mutually independent configurations
Boosting hyperspectral image classification with Gate-Shift-Fuse mechanisms in a novel CNN-Transformer approach
No full textInternational audienceHyperspectral Image (HSI) classification categorizes every pixel sample into distinct land-cover types. While CNN-based methods have demonstrated significant advances in feature representation, they often struggle with extracting deep features. On the other hand, transformer models excel in capturing high-level semantic features, offering complementary strengths. This paper introduces a novel HSI classification framework that combines convolutional blocks with Gate-Shift-Fuse (GSF) and transformer blocks to leverage CNNs for local feature extraction and transformers for long-range context modelling. The GSF block enhances the integration of spatial-spectral features, while an effective attention mechanism module refines feature extraction from HSI cubes. The proposed method was evaluated on four benchmark datasets (Indian Pines, Pavia University, WHU-Hi-LongKou, and WHU-Hi-HanChuan). It achieved state-of-the-art overall accuracies (OA) of 99.47% on the Indian Pines dataset and 99.86% on WHU-Hi-LongKou, demonstrating significant improvements over existing models. These results highlight the framework's ability to address challenges like spectral variability and unbalanced datasets. Our contributions include the integration of the GSF block for enhanced spatialspectral feature fusion and the application of transformers for robust high-level semantic understanding, setting a new benchmark in HSI classification
A multiscale approach to investigate the impact of thermal fluctuations on crack propagation in fracture and decohesion
No full textInternational audienceIn various physical, biological and technological systems, fracture propagation and decohesion phenomena are widespread. Understanding these processes has broad applications, ranging from classical solid mechanics (e.g., crack propagation) to emerging fields such as advanced materials, nanotechnology, and soft materials like rubber-like substances and biomaterials (e.g., cell adhesion and de-adhesion, DNA denaturation).Experimental findings have already highlighted the important role that thermal fluctuations can play on these phenomena. However, a theoretical investigation of temperature effects remains challenging due to the complexity of the involved physical processes. Thus, despite extensive experimental studies and numerical simulations investigating the interplay between thermal fluctuations and mechanical properties, a comprehensive theoretical framework fully integrating temperature effects is still lacking. In this study we present a first step in this direction, based on a simplified model for the Griffith energy criterion for mode I fracture, extending the classical approach to include thermal fluctuations. The key step in this direction lies in replacing the total mechanical energy by including entropicterms and referring to the free energies. Thuss, by employing tools from equilibrium statistical mechanics, we incorporate entropic effects into the overall energy balance.Furthermore, we adopt a multiscale paradigm by developing both discrete and continuum models. Although intentionally simplified, these models allow for analytical solutions and provide deeper insights into the underlying physics. Our approach effectively captures the energetic competition between elastic deformation, external loading, fracture energy, and entropic contributions.The proposed framework describes not only the scenario where fracture or decohesion propagates from one end but also cases where the broken region forms inside the system, including configurations with multiple fracture bubbles (e.g., multiple bubbles in DNA). In the latter case, we identify solutions where these bubbles merge before complete failure occurs.Finally, our model reveals a classical critical behaviour: the critical load decreases as temperature increases, following the law (1 − T /Tc)^(1/2) . As a direct consequence of temperature effects, we demonstrate that, at the critical temperature Tc , the system undergoes a phase transition, leading to complete rupture without any applied mechanical load
Room-temperature, continuous-wave terahertz generation in free-space with an intersubband mid-infrared photomixer
No full textInternational audienceWe demonstrate a terahertz photomixer pumped by mid-infrared photons at a wavelength of 10 μm. The device is based on a dual-antenna architecture, in which a two-dimensional array of mid-infrared patch-antenna resonators is connected to the electrodes of a log-spiral terahertz antenna. By exploiting intersubband absorption inside an AlGaAs–GaAs multi-quantum-well heterostructure, a photocurrent is coherently generated at the difference frequency between two quantum cascade lasers. The photocurrent drives the antenna electrodes allowing tunable, continuous-wave generation in free-space up to 1 THz at room temperature. By experimentally studying the photomixer frequency response and dark impedance, we demonstrate that the observed power roll-off at high-frequency is limited by the combined effect of the photoexcited carrier’s intrinsic lifetime and of the device RC time constant. In the spectral range investigated, we obtain, in continuous-wave, mid-infrared → terahertz conversion efficiencies exceeding by orders of magnitude those of unipolar devices exploiting χ(2) processes
