HAL-Université de Bretagne Occidentale
Not a member yet
75442 research outputs found
Sort by
Contributions au décodage neuronal des codes en blocs linéaires
No full textThis thesis addresses the challenge of soft-decision decoding for short block codes by investigating the integration of machine learning into decoding, considering both model-free and model-based paradigms. For model-free approaches, syndrome-based neural decoders (SBND) are studied to estimate error patterns and approximate maximum likelihood decoding (MLD). We show that fixed datasets, optimized sampling, and automorphism-based data augmentation significantly improve generalization and sample efficiency. When coupled with inference-time post-processing, the decoding performance approaches MLD. For model-based approaches, we consider multi-round belief propagation (MRBP) assisted by neural networks By leveraging SBND-inspired input representations and stacked recurrent architectures, we demonstrate that the problematic variable nodes causing BP failures can be identified more accurately, reducing the number of decoding attempts and outperforming classical expert-based rules. These results highlight the need for more synergy between structural code knowledge and data-driven learning, offering a promising avenue for near-optimal decoding of short block codes.Cette thèse aborde le défi posé par le décodage à décision souple pour les codes à blocs courts par le biais de l'intégration de l'apprentissage automatique dans le décodage, soit par l’augmentation d’algorithmes existants, soit par approche sans modèle. Pour les approches sans modèle, nous utilisons le décodage neuronal basé sur le syndrome pour estimer le motif d'erreur le plus vraisemblable et approcher ainsi le décodage optimal. Nous montrons que l’utilisation de jeux d’entraînement bien construits par échantillonnage optimisé des exemples, combinée à de l'augmentation de données basée sur les automorphismes du code améliore considérablement la capacité de généralisation du modèle et l’efficacité de l’entraînement. Cette approche peut être complétée par des post-traitements à l’inférencepour se rapprocher encore davantage de la performance cible du décodeur optimal. Pour les approches basées sur des modèles, nous considérons un décodage par propagation de croyance (BP) augmenté, et assisté par un réseau de neurones. En tirant profit des représentations d'entrée et architectures récurrentes développées pour le décodage neuronal par syndrome, nous montrons que les nœuds variables problématiques à l'origine des échecs du décodage BP peuvent être identifiés avec plus de précision, ce qui réduit le nombre de tentatives de décodage et surpasse les règles expertes classiques. Ces résultats mettent en évidence l’intérêt d’une synergie entre connaissance structurelle des codes et apprentissage basé sur les données, dessinant ainsi une voie prometteuse pour le décodage quasi optimal des codes courts
Feature-Guided Metaheuristic with Diversity Management for Solving the Capacitated Vehicle Routing Problem
No full textInternational audienceWe propose a feature-based guidance mechanism to enhance metaheuristic algorithms for solving the Capacitated Vehicle Routing Problem (CVRP). This mechanism leverages an Explainable AI (XAI) model to identify features that correlate with high-quality solutions. These insights are used to guide the search process by promoting solution diversity and avoiding premature convergence. The guidance mechanism is first integrated into a custom metaheuristic algorithm, which combines neighborhood search with a novel hybrid of the split algorithm and path relinking. Experiments on benchmark instances with up to 30,000 customer nodes demonstrate that the guidance significantly improves the performance of this baseline algorithm. Furthermore, we validate the generalizability of the guidance approach by integrating it into a state-of-the-art metaheuristic, where it again yields statistically significant performance gains. These results confirm that the proposed mechanism is both scalable and transferable across algorithmic frameworks
Articuler la théorie des pratiques et les sciences comportementales pour comprendre et renforcer le tri des déchets recyclables en habitat collectif
No full textInternational audiencePurpose Despite their environmental and economic importance, recycling rates remain low in collective housing. Public policies grounded in decontextualized psychological mechanisms show limited effectiveness in this setting. By contrast, practice-based research shifts attention to contextualized everyday routines but offers few operational pathways for intervention. This study aims to advance a more integrated perspective by examining how psychological mechanisms are activated, inhibited or reshaped by collective housing through residential elements of materiality, meanings and competences. Design/methodology/approach Twenty-two semi-structured interviews were conducted with residents living in collective housing in Paris and the surrounding departments. Psychological mechanisms were explored in depth through the practice elements of materiality, meanings and competences. Findings The results show that the residential context deeply shapes how individuals engage with waste sorting. Affordances and emotions elicited by shared spaces profoundly shape sorting practices. Residents actively reconstruct residential sorting norms and cooperation expectations, and experience shame, guilt and pride tied to recycling. Competences extend beyond the household, shaping how residents navigate shared spaces, communicate with neighbors and restore order within the residence. Research limitations/implications Future research should further examine how residential context shapes consumption practices within a systemic perspective. Observational methods, such as sensor data or resident-generated video, could help identify key change points in daily routines. Experimental designs could test residential mechanisms, such as how residential altruistic acts shape waste-sorting behavior. Finally, specialized tools like causal loop diagrams may support a systemic approach to the dynamics of social and consumption change. Practical implications This study offers a framework for field diagnostics and context-based interventions that target the residential environment rather than individuals, focusing on residential psychological mechanisms. Originality/value From a critical realist perspective, collective housing is not simply a constraint on change but a generative structure that shapes waste-sorting routines by activating and shaping residential psychological mechanisms
Clicking [Mo(CO) 3 (triphos)] to an azide-terminated alkylthiol SAM: structural, spectroscopic and electrochemical investigations
No full textInternational audienceThe synthesis and characterization of a new, linear triphos-type pincer ligand furnished with an ethynyl moiety in the ligand backbone (etPP PhTMS P) and the derived tricarbonyl complex [Mo(CO) 3 (etPP PhTMS P)] as well as its covalent attachement onto an Au-surface are described
'Desacralizing Joy: The Travel Writings of Two British Female Loyalists in Revolutionary France
No full textInternational audienceTravel accounts of revolutionary France by conservative British women were used as valuable sources by nineteenth-century French historians to document the social effects of the Revolution. Since then, they have received little attention. This paper focusses on the writings of Rachel Charlotte Biggs (c.1763–1827) and Louise Elisa Beaumont (1751–1818), who both traveled to France to offer eyewitness accounts, albeit with a strong loyalist bias, of the extraordinary social plight of a people caught up in the most violent political commotion modern Europe had ever seen. They chronicled the everyday life of the newly founded republic, thus offering a sociology of revolutionary change, with an eye to collective emotions and moral effects. Their loyalist accounts substantiated the predictions of Edmund Burke, who had foretold the fate of France in Reflections on the Revolution in France. The chilling and claustrophobic world they depict, locked up in an indefinite present with neither hope for the future nor a sense of utopia to strive for, was the outcome of the revolutionary transformation of society, which had alienated the individual from his social and spiritual matrix
Anisotropic fluorescence emission by diatoms modifies the underwater oceanic light field
No full textInternational audienceFluorescence in phytoplankton and other autotrophic organisms originates within the cell chloroplasts, where a fraction of the absorbed solar radiation is reemitted at longer wavelengths by photopigments. While traditionally employed as an indicator of physiological status, emerging evidence suggests that natural chlorophyll fluorescence (ChlF) may also play unanticipated functional roles in the marine environment. Here, we examine the ChlF emission fields generated by pennate planktonic diatoms, a key phytoplankton group playing a critical role in global biogeochemical cycles. Using cell micromanipulation experiments, we demonstrate that the ChlF emitted by Pseudo-nitzschia fraudulenta (Bacillariophyceae) is markedly anisotropic, a feature attributed to both the cells’ elongated morphology and the arrangement of chloroplasts within the cytoplasm. In these diatoms, fluorescence is preferentially emitted in the transapical direction, accounting for up to 35% of total emission. However, peak ChlF intensities occur at the cell apices, suggesting that the silica frustule focuses fluorescent light emission along the longitudinal axis. At elevated cell densities (~10 6 cells/L), the underwater light field is modulated by the combined effects of ChlF emission anisotropy and preferential alignment of diatom cells within the water column. Numerical simulations indicate that ChlF intensity can vary by up to 15% depending on whether cells are predominantly aligned—commonly in stratified water columns—or randomly oriented. These results suggest that diatom-driven modulation of the light field through structured ChlF emission may influence microscale optical environments, with potential consequences for processes ranging from intercellular signaling to large-scale phytoplankton dynamics, including remote sensing–based assessments of phytoplankton physiology
Production of 13C-labeled docosahexaenoic acid from heterotrophic marine microorganisms Aurantiochytrium mangrovei and Crypthecodinium cohnii enabling fluxomic applications
No full textInternational audienceDocosahexaenoic acid (DHA, 22:6n-3) is the predominant polyunsaturated fatty acid in the human brain and eyes, playing a crucial role in vision and cognitive development. DHA deficiency has been associated with ocular diseases, such as macular degeneration and glaucoma, as well as neurodegenerative disorders. Since the human body has a limited ability to synthesize DHA from its precursor, alpha-linolenic acid (ALA, 18:3n-3), targeted DHA supplementation is essential for these patients. To investigate DHA metabolism and integration, researchers commonly use stable ( 2 H, 13 C) or radioactive ( 3 H, 14 C) isotopes, which are expensive and not widely accessible, restricting the scope and duration of studies. This study aimed to develop a sustainable method for biosynthesizing uniformly labeled 13 C-DHA by culturing the heterotrophic protists Aurantiochytrium mangrovei and Crypthecodinium cohnii with 13 C-glucose. The major fatty acids (FA) of A. mangrovei included 16:0, 22:5n-6 (DPA), and DHA, with DHA accounting for 50.5% ± 4.9% of the total FA. Gas Chromatography-Mass Spectrometry (GC-MS) analysis revealed a 13 C-enrichment of DHA at 96.7% ± 0.4% after the effective High Performance Liquid Chromatography (HPLC) purification. The predominant FA of C. cohnii were 12:0, 14:0, 16:0, and DHA, with DHA representing around 27% of the total FA and exhibiting a 13 C-enrichment of 86.3% ± 1.6%. Based on FA content, A. mangrovei showed a balanced distribution of neutral and polar lipids, with DHA predominantly in the polar fraction (57.8% ± 3.1%), whereas C. cohnii exhibited a predominance of neutral lipids (82.4% ± 0.3%), which contained the majority of its DHA (57.5% ± 1.0%)
Recycling or downcycling? The difference between raw and stabilised earth explained
No full textInternational audienceThe development of innovative earthen materials necessitates a thorough evaluation of their recyclability. Although stabilisation is often regarded as a barrier to the recyclability of earthen materials, this study investigates the feasibility of recycling compressed earth blocks (CEBs) stabilised with 5% CEM III A or NHL 3.5. To evaluate this property, the study involved repeated cycles of material characterization, compaction using the modified Proctor method, compressive strength testing, water durability tests and subsequent crushing. The raw material remained unchanged through the recycling process, retaining its clay activity, manufacturing properties, and compressive strength. In contrast, the stabilised material showed a significant decline in clay activity, rendering it negligible. After three recycling cycles, the Methylene Blue Value (MBV) of the cement- and lime-stabilised materials decreased from 0.7 to 0.1, while the materials do not show any plastic behaviour whatever their water content. Furthermore, the use of these stabilisers resulted in a decrease in the optimal dry density by 80 kg/m 3 for CEM III A and 50 kg/m 3 for NHL 3.5. Despite this reduction, there was no corresponding decrease in compressive strength following the first recycling cycle. This indicates the binder's effectiveness may remain similar because of the diminished clay activity. These results also indicate that stabilised materials must be re-stabilised during recycling to prevent a loss of compressive strength compared to the raw material. Considering these experimental results, recycling stabilised earthen materials should rather be defined as a downcycling process or a recycling process that needs additional stabiliser while raw earth displays an infinite recyclability
Coastal winds are stronger and impact seabird behaviour and fitness in a warming Arctic
No full textThe Arctic is warming nearly four times faster than the rest of the planet. This unprecedented warming is profoundly altering environmental conditions. Coastal areas, which are crucial for biodiversity and human activities, are at the forefront of these changes. Among the major environmental changes expected in coastal regions are shifts in wind regimes and an increase in storm frequency. Using an interdisciplinary approach, we demonstrated that (1) wind conditions in Arctic coastal regions during summer are changing, resulting in a rise in strong wind events. (2) These large-scale changes in wind dynamics influence the movements and fitness of seabirds, with broader impacts on panarctic socio-ecosystems. To demonstrate these effects, we used a long-term study of little auks (Alle alle), the most abundant Arctic seabird species and an ecological indicator of Arctic coastal ecosystems. We showed that during strong wind events, little auks significantly reduce the visits to the colony. These behavioural changes negatively impact their fitness as inferred from chick growth rates. Overall, strong wind events disrupt the delicate balance between parental investment and self-maintenance in this long-lived species. Over time, such pressures may drive shifts in life-history traits, potentially affecting population dynamics