HAL-IRD
Not a member yet
83947 research outputs found
Sort by
EM-SEC: Efficient Multi-head Set-valued Evidential Classification
No full textInternational audienceIn machine learning and deep learning, uncertainty quantification helps to accurately assess a model's confidence in its predictions, enabling the rejection of uncertain outcomes in safety-critical applications. However, in scenarios involving AI-assisted decision-making, proposing multiple plausible decisions can be more beneficial than either not making any decisions or risking incorrect ones. Set-valued classification is a relaxation of standard multiclass classification where, in cases of uncertainty, the classifier returns a set of potential labels instead of a single label. Current methods for set-valued classification often suffer from high computational complexity or fail to adequately quantify uncertainty. In this paper, we introduce a novel, computationally efficient approach to set-valued classification leveraging evidential deep learning and subjective logic, explicitly providing a measure of classification uncertainty. Our method employs a dual-head architecture: one head conducts multiclass evidential classification, while the other suggests candidate label sets when uncertainty is high. The proposed approach has linear worstcase computational complexity with respect to the number of classes. Extensive evaluation on several benchmark datasets demonstrates that our method showcases comparable performance to baseline set-valued methods, while being up to 23 times faster at inference on the benchmark datasets
Explorer les performances de l'agroforesterie à travers l'analyse décisionnelle multi-critères d'aide à la décision (AMCD) : enseignements tirés d'une évaluation participative dans le sud-ouest de la France
No full textInternational audienceIn a context of ecological degradation, climate instability, and soil exhaustion, supporting agroecological transitions requires not only agronomic innovation but also decision-support tools capable of handling complex, multi-dimensional systems. This work presents the use of Multiple Criteria Decision Aiding (MCDA)—specifically ELECTRE Tri-C and ELECTRE III—to evaluate the performance of agroforestry systems (AFS) in southwestern France.Conducted within the AC²TION project, the study involves 29 farmers and 33 AFS plots. A participatory process led to the co-construction and weighting of seven socio-agro-environmental criteria: functional biodiversity, resilience of agronomic practices, landscape amenities, additional workload, care of trees and shrubs, health of ligneous plants, and access to knowledge and guidance.As the AFS are still young, their measurable impact on soil and climate remains limited. We therefore developed a separate “Soil Potential” (PSOL) criterion—based on biophysical soil characteristics—as an explanatory variable used post hoc to interpret performance differences revealed by the MCDA. Climate factors were treated similarly.Our results show that MCDA can structure stakeholder knowledge and diverse indicators into actionable insights, even in systems marked by uncertainty and limited historical data. This approach supports better-informed public policy and helps identify potential levers or barriers to the agroecological transition.In line with the theme “From Memories to the Future”, this work explores how integrating ecological memory—soil characteristics and long-term system functions—with current stakeholder knowledge can inform the development of agroforestry, a promising but still marginal practice in the region.Dans un contexte de dégradation écologique, d'instabilité climatique et d'épuisement des sols, soutenir les transitions agroécologiques nécessite non seulement des innovations agronomiques, mais aussi des outils d'aide à la décision permettant de gérer des systèmes complexes et multidimensionnels. Cette étude présente l'utilisation de méthode multicritères d'aide à la décision (AMCD), en particulier ELECTRE Tri-C et ELECTRE III, pour évaluer les performances des systèmes agroforestiers (SAF) dans le sud-ouest de la France.Réalisée dans le cadre du projet AC²TION, l'étude porte sur 29 agriculteurs et 33 parcelles agroforestières. Un processus participatif a permis de co-construire et de pondérer les critères socio-agro-environnementaux : biodiversité fonctionnelle, résilience des pratiques agronomiques, aménagements paysagers, charge de travail supplémentaire, entretien des arbres et arbustes, santé des plantes ligneuses, et accès aux connaissances et aux conseils.Les SAF étant encore jeunes, leur impact mesurable sur les sols et le climat reste limité. Nous avons donc développé un critère distinct, le « potentiel des sols » (PSOL), basé sur les caractéristiques biophysiques des sols, que nous utilisons a posteriori comme variable explicative pour interpréter les différences de performance révélées par l'AMCD. Les facteurs climatiques ont été traités de la même manière.Nos résultats montrent que l'AMCD permet de structurer les connaissances des parties prenantes et les différents indicateurs en informations exploitables, même dans des systèmes caractérisés par l'incertitude et des données historiques limitées. Cette approche permet d'élaborer des politiques publiques plus éclairées et d'identifier les leviers ou les obstacles potentiels à la transition agroécologique.Conformément au thème « Des souvenirs vers l'avenir » de la conférence, ce travail explore la manière dont l'intégration de la mémoire écologique (caractéristiques du sol et fonctions à long terme du système) aux connaissances actuelles des parties prenantes peut éclairer le développement de l'agroforesterie, une pratique prometteuse mais encore marginale dans la région
Pixel-based satellite mapping for coral island seabed classification: Application to the Maupiti island, French Polynesia
No full textInternational audienceThis study introduces a novel pixel-based satellite mapping approach for classifying coral island seabed. The model combines a pixel-based approach and a segmentation technique, to smooth the predictions into coherent objects. The model is applied to Maupiti Island (French Polynesia) and compared with an expert-based mapping mostly based on the Reef Cover classification. Results demonstrate high accuracy, ranging between 87% and 90% for various spatial resolutions. The developed tool is open-source and flexible, allowing users to retrain it for different classification schemes and environments. The study highlights the potential of automated satellite mapping for monitoring coral reef ecosystems and supporting conservation efforts
Combining transient dynamics and logistic‐asymptotic growth to study the recovery of two seabird populations after rat eradication
No full textInternational audienceUnderstanding demographic processes is crucial in ecology and conservation biology to assess how populations respond to environmental pressures. Although demographic parameters often react to disturbances, populations may remain in unstable age structures (transient dynamics) following such events. To explore the influence of transient dynamics in masked booby and red‐footed booby populations on Tromelin Island 15 years after rat eradication, we compared observed, transient, and asymptotic growth using count and capture–recapture (CR) data. CR data were used to estimate survival probability from Cormack–Jolly–Seber models and other vital rates such as breeding frequency and age at first breeding. We coupled asymptotic growth with a density‐dependent model for breeding site availability to project long‐term population growth. Both populations exhibited high adult survival (0.91–0.96) and breeding success (0.46–0.54). The masked booby juvenile survival was also high (0.86). Populations are expected to grow until breeding sites are saturated, in about a century. Comparison of asymptotic and transient growth suggests that population growth is linked to an intrinsic process and that both populations have undergone disruption in their age structure likely due to past predation by rats. Although the observed population growth appears to align more closely with transient growth, we could not entirely exclude the possibility that differences between asymptotic and observed growth rates may be due to data scarcity or age‐related variation in demographic parameters. Nevertheless, the study of transient dynamics allowed a better understanding of the differences observed between field counts and estimated asymptotic growth rates based on matrix population models
In situ U–Pb dating of upper triassic magmatic flows (Spilite) in the External Western Alps (Pelvoux massif): A peripheral CAMP activity?
No full textInternational audienceMagmatic activity during the Triassic-Jurassic transition coincided with the breakup of Pangea, marking a pivotal period in Western European tectonic evolution. However, this activity remains poorly documented in the External Western Alps. Dating the emplacement and alteration of Triassic magmatic rocks such as the spilites in the Alps has long been challenging due to the complex alteration history and the scarcity of suitable mineral phases. This study employs in situ U-Pb dating of carbonate to constrain the timing of hydrothermal alteration of spilites from the Pelvoux massif (France), offering a new temporal framework for these processes. Dolomite and calcite filling vesicles and veins in the spilites yield ages of 201 ± 15 Ma and 202 ± 47 Ma, respectively, consistent with the stratigraphic emplacement interval of these lavas during the Upper Triassic. Notably, the U-Pb system in dolomite has preserved hydrothermal conditions related to magmatic emplacement, without resetting during two subsequent geological events at temperatures approaching 300°C, emphasizing its reliability as a chronometer in this context. The obtained ages overlap with the temporal framework of the Central Atlantic Magmatic Province (CAMP) and geochemical signatures of the spilites correspond to medium/high-Ti transitional to alkaline basalts, comparable to continental basalts such as those of the CAMP. This suggests a shared tectono-magmatic context with mantle-derived magmatism. Furthermore, the spatial proximity of the studied spilites to lower crustal CAMP-related magmatism in the Internal Alps supports a potential genetic relationship, with magma ascent likely facilitated by inherited tectonic structures during the Upper Triassic extension and the opening of the Alpine Tethys. Hydrothermal alteration, marked by spilitization and carbonate precipitation, occurred under low- to moderate-temperature conditions (70–360°C), possibly driven by marine or continental-derived fluids. By providing the first absolute geochronological constraints on spilites in the External Western Alps, this study expands the recognized extent of CAMP. It underscores the utility of carbonates as reliable archives for unraveling hydrothermal and magmatic histories
Tuning the Diameter of Supramolecular Nanocylinders: Balancing Long and Short Polymer Arms for Optimized Self‐Assembly
No full textInternational audienceNanocylinders find applications in biology, catalysis, and material science due to their high aspect ratio and nanometric diameter, granting them a high specific area. They can be formed by supramolecular assembly in solution of polymers decorated by stickers promoting directional assembly. Controlling the dimensions of nanocylinders impacts their applicative properties. Many strategies exist to tune their length, but much fewer to tune their diameter. We address this point here by synthesizing water‐soluble polymers of different lengths end‐functionalized by hydrogen bonding tris(urea) stickers. It is shown through a combination of light/neutron scattering experiments and cryo‐transmission electron microscopy that short polymer arms (DP ≤ 80) produce thin (d = 10 nm) and long (L > 500 nm) nanocylinders, whereas longer polymer arms (DP ≥ 500), required to significantly increase d, inhibit self‐assembly due to the strong entropic penalty caused by their stretching. A compromise to control d is to mix a few long arms, which increase d, with a sufficient amount of short arms, which alleviate the entropic penalty and maintain high L. This approach proposes a straightforward way to tune the specific area of nanocylinders—a key parameter for applications in catalysis, emulsion stabilization, or interactions with biological materials
Closed-loop recycling of bio-based disulfide vitrimer via a solvent-and waste-free strategy
No full textInternational audienceA solvent-free process enables the efficient degradation of the vitrimer matrix under mild conditions, producing oligomers that can be reused without purification
Direct assembly of micrometer-long polymeric cylinders in water via supramolecular sticker engineering
No full textInternational audienceWe report a direct, solvent-free method to produce micrometer-length, well-organized polymer nanocylinders in water. To achieve this, a hydrophilic poly(N,N-dimethylacrylamide) (PMDAc) was functionalized at one chain-end with a perylene diimide (PDI) sticker using RAFT polymerization. Two PDI RAFT agents were prepared and studied: one featuring two tri(ethylene glycol) (TEG) units at the PDI bay-positions and one without. The corresponding PDI-PDMAc conjugates spontaneously self-assemble in water, driven by π-π interactions made of H-aggregates, and show a morphological evolution from cylinders to spheres when increasing the polymer chain length. The introduction of TEG units was found to be important to avoid the clustering of nanocylinders or the formation of ill-defined assemblies, which were observed in the TEG-free system. Moreover, we found that the PDI-TEG 2 -PDMAc with degrees of polymerization (DP n ) below 24 self-assembled into micrometer-long nanocylinders. By heating the aqueous polymer solution, this process can be accelerated and is accompanied by a large increase in viscosity. Fluorescence spectroscopy revealed an excimer emission signal for the PDI polymers in water, with a higher emission for cylinders, suggesting better organization within the PDI H-aggregates. This strategy provides a sustainable approach for developing functional nanomaterials with precise morphological control, eliminating organic solvents and complex processing
Alkali-promoted FeCo/TiO<sub>2</sub> catalysts for carbon dioxide hydrogenation to fuels
No full textInternational audienceA titania support has been modified in order to prepare alkali promoted cobalt-iron catalysts for the hydrogenation of carbon dioxide into long-chain hydrocarbons (C5+). Under mild reaction temperatures (220-250-270 °C), the partial substitution of cobalt by iron decreased the catalytic activity (CO2 conversion) but increased the selectivity into C5+, since iron is less active within this temperature range, but more selective for the reverse water-gas shift reaction. Alkali promotion (Na or K) further decreased the catalytic activity of cobalt-iron catalysts, albeit, strongly increasing the selectivity into CO and C5+ by decreasing CH4 production. This could be attributed to the high basicity of alkali promoted catalysts, their beneficial effect on CO2 dissociative adsorption, and their ability to limit H2 adsorption
Magneto-chiral dichroism at the nanoscale: experimental observation in chiral paramagnetic nanoparticles
No full text"ADC - Accord Couperin / American Chemical Society (2024-2026)"International audienceHerein we report the first observation of magneto-chiral dichroism (MChD) on paramagnetic nanoparticles. Magnetic nanoparticles based on cobalt(II) were synthesized in the presence of enantiopure (D or L)-aspartic acid. The chiral ligands coordinate the Co 2+ ions at the nanoparticle surface inducing a strong natural circular dichroism (gNCD = 10 -2 ) for the 4 T1( 4 P) ← 4 T1( 4 F) electronic transitions of the Co 2+ ion in octahedral sites. Magneto-chiral dichroism (MChD) measurements reveal an unambiguous MChD signal at low temperatures for the same transitions, which represent the first evidence of MChD for functionalized nanoparticles. These results demonstrate the possibility of detecting the magneto-optical properties of a nanomaterial without the need for light polarization and open alternative perspectives in magnetic sensing, optical devices, and biomedical fields.</p