Portail HAL de l'Université du Littoral Côte d'Opale
Not a member yet
11255 research outputs found
Sort by
Ergodicity of an Adaptive MCMC Sampler under a Probability Bound
No full textThis paper provides sufficient conditions over the sequence of samples and parameters of an adaptive Markov Chain Monte Carlo (MCMC) algorithm to ensure ergodicity with respect to a target distribution that can have unbounded support. These conditions aim to make more easily usable the conditions of Containment and Diminishing Adaptation from Roberts and Rosenthal [2007] formulated over the transition kernels, without needing, as was done in other works, an artificial assumption of the compactness over both sample and parameter spaces. The paper shows that the condition of compactness can be relaxed to a more realistic bound in probability over the sequence of both samples and parameters
A New Combination of Preconditioned Gradient Descent Methods and Vector Extrapolation Techniques for Nonlinear Least-Squares Problems
No full textVector extrapolation methods are widely used in large-scale simulation studies, and numerous extrapolation-based acceleration techniques have been developed to enhance the convergence of linear and nonlinear fixed-point iterative methods. While classical extrapolation strategies often reduce the number of iterations or the computational cost, they do not necessarily lead to a significant improvement in the accuracy of the computed approximations. In this paper, we study the combination of preconditioned gradient-based methods with extrapolation strategies and propose an extrapolation-accelerated framework that simultaneously improves convergence and approximation accuracy. The focus is on the solution of nonlinear least-squares problems through the integration of vector extrapolation techniques with preconditioned gradient descent methods. A comprehensive set of numerical experiments is carried out to study the behavior of polynomial-type extrapolation methods and the vector ε-algorithm when coupled with gradient descent schemes, with and without preconditioning. The results demonstrate the impact of extrapolation techniques on both convergence rate and solution accuracy, and report iteration counts, computational times, and relative reconstruction errors. The performance of the proposed hybrid approaches is further assessed through a benchmarking study against Gauss--Newton methods based on generalized Krylov subspaces
Beyond the prompt : gender ratio in text-to-image models, with a case study on hospital professions
No full textInternational audienceText-to-image (TTI) models are increasingly used in professional, educational, and creative contexts, yet their outputs often embed and amplify social stereotypes. This paper investigates gender representation in six state-of-the-art open-weight models: HunyuanImage 2.1, HiDream-I1-dev, Qwen-Image, FLUX.1-dev, Stable-Diffusion 3.5 Large, and Stable-Diffusion-XL. Using carefully designed prompts, we generated 100 images for each combination of five hospital-related professions (cardiologist, hospital director, nurse, paramedic, surgeon) and five portrait qualifiers ("", corporate, neutral, aesthetic, beautiful). Our analysis reveals systematic occupational stereotypes: all models produced nurses exclusively as women and surgeons predominantly as men. However, differences emerge across models: Qwen-Image and SDXL enforce rigid male dominance, HiDream-I1-dev shows mixed outcomes, and FLUX.1-dev skews female in most roles. HunyuanImage 2.1 and Stable-Diffusion 3.5 Large also reproduce gender stereotypes but with varying degrees of sensitivity to prompt formulation. Portrait qualifiers further modulate gender ratio, with terms like corporate reinforcing male depictions and beautiful favoring female ones. Sensitivity varies widely: Qwen-Image remains nearly unaffected, while FLUX.1-dev, SDXL, and SD3.5 show strong prompt dependence. These findings demonstrate that gender stereotype in TTI models is both systematic and model-specific. Beyond documenting disparities, we argue that prompt wording plays a critical role in shaping demographic outcomes. The results underscore the need for bias-aware design, balanced defaults, and user guidance to prevent the reinforcement of occupational stereotypes in generative AI
When is Cat(Q) Cartesian Closed?
No full textInternational audienceWe give an elementary characterization of those quantaloids Q for which the category Cat(Q) of Q-enriched categories and functors is cartesian closed. We then unify several known cases (previously proven using ad hoc methods) and we give some new examples
Ultrafiltration-based recovery of kraft lignin from black liquor for depolymerization purposes
No full textInternational audienceLignin, the most abundant aromatic biopolymer on Earth, represents a promising renewable resource for sustainable production of bio-based chemicals and energy carriers. However, in the pulp and paper industry, the majority of the 70 million tons of lignin annually generated in kraft black liquor remains underutilized due to its complex composition. This study explores ultrafiltration (UF) as a purification strategy to recover lignin fractions suitable for enzymatic depolymerization. Softwood black liquor was treated using six polymeric membranes (polyethersulfone, regenerated cellulose) with molecular weight cut-offs of 30, 10, and 5 kDa. UF achieved high lignin retention (>95%) across all membranes, while simultaneously rejecting 70–80% of sugars, indicating the presence of lignin–carbohydrate complexes (LCCs). Structural analyses by 31 P NMR and gel permeation chromatography confirmed that LCCs contributed to elevated apparent molecular weights and influenced membrane performance. An acid precipitation pre-treatment cleaved ether-linked carbohydrates, reducing lignin molecular weight by ~40% and improving UF selectivity. Among the tested configurations, the 5 kDa PES membrane proved most effective, combining high retention with favorable permeability and flux recovery after cleaning. These findings provide new insights into LCC-driven separation mechanisms and identify UF conditions enabling the production of purified lignin fractions compatible with enzymatic depolymerization, thereby supporting the development of integrated and sustainable lignin valorization routes in future biorefineries
Unlocking the potential of recycled waste cooking oil for a sustainable volatile organic compound absorption process
No full textInternational audienceFinding suitable absorbents to remove volatile organic compounds (VOCs) from industrial polluted air remains apressing scientific and industrial challenge. Most VOCs are hydrophobic and water demonstrates limited masstransfer in this case. In this study, we evaluated the feasibility of using recycled waste cooking oil (RWCO) toabsorb VOCs for the first time. We compared the performance of RWCO with that of two deep eutectic solvents(DESs), including a hydrophobic DES prepared from decanoic and dodecanoic acids (C10:C12) and a supramolecular DES based on randomly methylated β-cyclodextrin (RAMEB) and levulinic acid, and with a conventional solvent, propylene glycol (PG), in absorbing four hydrophobic VOCs, toluene, limonene, siloxane D4, anddecane. The vapor-liquid partition coefficients (K) of the VOCs alone and in mixture were determined using staticheadspace gas chromatography (SH-GC). A lab-scale bubbling device was used to measure absorption capacitiesunder simulated industrial conditions. Results from the static method showed that RWCO performed similarly tothe hydrophobic DES C10:C12, exhibiting the highest absorption affinities for the tested VOCs, with K values upto 208 000 times lower than in water. Dynamic absorption studies corroborated these results, highlighting RWCOand C10:C12 as the most effective absorbents, with an absorption capacity up to 0.884 mg/g for toluene at 100ppm. This study also demonstrated that bubbling nitrogen in addition to heating at 60◦C reduced regenerationtime from 48 hours to 2.5 hours. The RWCO and C10:C12 absorbents retained over 99.5% of their initial absorption capacity after undergoing twelve consecutive absorption-desorption cycles. These results suggest thatRWCO could be an effective and sustainable absorbent for VOCs abatement while complying with circulareconomy concept
Toward sustainable transformer oil recycling: comparative efficacy of adsorbents and CFD simulation analysis
No full textInternational audienceTransformer insulating oils undergo progressive degradation due to oxidation, accumulation of aging by-products, and contamination by metallic particles. Proper regeneration of these oils is crucial to extend transformer lifespan and reduce environmental impact. This study investigates the performance of locally activated Algerian adsorbents (Illite and Maghnite) compared to commercial adsorbents (activated Bauxite and Sepiolite) for the regeneration of degraded transformer oils. The materials were evaluated for their adsorption efficiency toward acids, water, and oxidation products, as well as their ability to restore key electrical and thermalproperties. Activated Illite and Maghnite demonstrated performance comparable to commercial adsorbents, significantly reducing the acid number from 0.23 to 0.01 mg KOH/g. Illite exhibited superior reusability, maintaining regeneration efficiency over 300 cycles. Additionally, sulfur compounds were reduced from 344 to 40 ppm. Treated oils achieved a breakdown voltage of 78 kV and a dielectric dissipation factor of 0.995, closely matching new oil standards. Computational fluid dynamics (CFD) simulations were employed to assess Illite’s behavior in industrial-scale oil treatment, supporting its effectiveness as a substitute for Bauxite. The results highlight Illite’s potential as a viable, sustainable, and locally available alternative for transformer oil recycling. Nevertheless, challenges related to large-scale implementation, including material availability, cost, and process integration, remain and warrant further investigation. This study contributes to the development of environmentally responsible and economically competitive solutions for the regeneration of insulating oils using natural adsorbents.Les huiles isolantes des transformateurs subissent une dégradation progressive due à l'oxydation, à l'accumulation de sous-produits de vieillissement et à la contamination par des particules métalliques. Une régénération adéquate de ces huiles est essentielle pour prolonger la durée de vie des transformateurs et réduire leur impact sur l'environnement. Cette étude examine les performances d'adsorbants algériens activés localement (illite et maghnite) par rapport à des adsorbants commerciaux (bauxite activée et sépiolite) pour la régénération d'huiles de transformateur dégradées. Les matériaux ont été évalués en fonction de leur efficacité d'adsorption des acides, de l'eau et des produits d'oxydation, ainsi que de leur capacité à restaurer les propriétés électriques et thermiques essentielles. L'illite et la maghnite activées ont démontré des performances comparables à celles des adsorbants commerciaux, réduisant considérablement l'indice d'acide de 0,23 à 0,01 mg KOH/g. L'illite a montré une réutilisabilité supérieure, conservant son efficacité de régénération pendant plus de 300 cycles. De plus, les composés soufrés ont été réduits de 344 à 40 ppm. Les huiles traitées ont atteint une tension de claquage de 78 kV et un facteur de dissipation diélectrique de 0,995, ce qui correspond étroitement aux nouvelles normes applicables aux huiles. Des simulations de dynamique des fluides computationnelle (CFD) ont été utilisées pour évaluer le comportement de l'illite dans le traitement du pétrole à l'échelle industrielle, confirmant son efficacité en tant que substitut à la bauxite. Les résultats soulignent le potentiel de l'illite en tant qu'alternative viable, durable et disponible localement pour le recyclage des huiles de transformateur. Néanmoins, les défis liés à la mise en œuvre à grande échelle, notamment la disponibilité des matériaux, le coût et l'intégration des processus, demeurent et méritent d'être approfondis. Cette étude contribue au développement de solutions respectueuses de l'environnement et économiquement compétitives pour la régénération des huiles isolantes à l'aide d'adsorbants naturels
Comparative Analysis of Lavandula Dentata Rhizosphere Microbiota Across Different Developmental Stages in a Semi‐Arid Area
No full textInternational audienceThe positive effects of soil microbiota on plant growth and stress tolerance are well established. However, their role in aromatic and medicinal plants, particularly under arid conditions, remains underexplored. This study examined rhizospheric microbial community dynamics across developmental stages of wild Lavandula dentata L . , a semi‐arid species threatened with extinction in Morocco. Results showed total microbial biomass peaked at senescence, mainly due to increases in Gram‐negative (25.02 μg/g) and Gram‐positive (18.11 μg/g) bacterial biomasses. Beta diversity analysis revealed consistent dominance of Actinobacteria, with peaks during senescence and the vegetative phase. Saprotrophic fungi (8.81 μg/g) and arbuscular mycorrhizal fungi (AMF) (4.16 μg/g) biomasses peaked at flowering. The fungal community was dominated by the Ascomycota phylum, with no significant variation across stages. The AMF genus Glomus remained most abundant throughout development. Senescence featured the most complex interkingdom interaction network and high ecological niche heterogeneity, reflected by more negative associations. Overall, the rhizospheric microbial community of L. dentata shifts with plant development, with flowering and senescence as key phases for microbial biomass accumulation and community diversification. Flowering and senescienceez stages seem to represent promising targets for developing biostimulant consortia to improve soil health and crop productivity in arid environments
Approche effective de l'algèbre tridendriforme des arbres de Schroeder
No full textWe introduce a primitive computation problem in the free tridendriform algebra generated by one element which is a Hopf algebra based on Schroeder trees. We know a complex way to generate all of them. To understand it clearer, we want to implement this method on a computer. However, we need to create some tools to implement Schroeder trees and the multiplications over this algebra to be able to compute the primitive elements. We also checked numerically that they are all primitive elements. In this paper, we detail how we made the problem mathematically understandable for a computer and how we implement it.Nous introduisons un problème de calcul des éléments primitifs dans l'algèbre tridendriforme libre des arbres de Schroeder générée par un élément munie de sa structure d'algèbre de Hopf. Une méthode compelxe permet de les générer. Cependant, elle n'est pas très bien comprise. Voilà pourquoi nous souhaitons implémenter cet algorithme. Pour cela, nous introduisons un point de vue sur les arbres permettant de les coder ainsi que les produits nécessaires et les coupes. Nous avons également vérifié numériquement que les résultats donnent des éléments primitifs. Dans ce document, nous détaillons le problème termes mathématiques compréhensibles pour un ordinateur