Portail HAL-PSL
Not a member yet
293219 research outputs found
Sort by
NKCC1 as a signaling hub regulating KCC2 stability, chloride homeostasis, and seizure susceptibility
No full textAbstract Chloride homeostasis relies on the dynamic balance between the neuronal co-transporters NKCC1 and KCC2. We reveal an unexpected mechanism by which NKCC1 governs KCC2 membrane stability. NKCC1 clusters recruits SPAK and PP1 to dynamically trap KCC2, compensating for its lack of a direct SPAK-binding site. Single-particle tracking shows that these NKCC1-rich assemblies operate as signaling hubs, enabling either SPAK-driven KCC2 phosphorylation and its membrane destabilization or PP1-mediated dephosphorylation of SPAK and KCC2 membrane stabilization. Peptides that activate SPAK by engaging NKCC1’s PP1-binding motif lower KCC2 surface levels and reduce chloride extrusion, whereas a SPAK-inhibiting peptide prevents SPAK recruitment to NKCC1, stabilizes KCC2 in membrane clusters, and enhances chloride extrusion. An optimized peptide analog preserves KCC2 clustering under hyperexcitable conditions, reduces seizure frequency and severity in PTZ-induced epilepsy, and suppresses ictal activity in human epileptic tissue. These findings identify NKCC1-KCC2 coupling as a central regulatory axis for inhibitory signaling, and position our peptides as promising therapeutic candidates to restore chloride homeostasis in epilepsy and other disorders marked by impaired KCC2 membrane stability
Unbiased Approximate Vector-Jacobian Products for Efficient Backpropagation
No full textIn this work we introduce methods to reduce the computational and memory costs of training deep neural networks. Our approach consists in replacing exact vector-jacobian products by randomized, unbiased approximations thereof during backpropagation. We provide a theoretical analysis of the trade-off between the number of epochs needed to achieve a target precision and the cost reduction for each epoch. We then identify specific unbiased estimates of vector-jacobian products for which we establish desirable optimality properties of minimal variance under sparsity constraints. Finally we provide in-depth experiments on multi-layer perceptrons, BagNets and Visual Transfomers architectures. These validate our theoretical results, and confirm the potential of our proposed unbiased randomized backpropagation approach for reducing the cost of deep learning
Development of a phase transient test bench to optimize a cold-atom microwave clock
No full textInternational audienceMuClock is a commercial transportable microwave clock developed at Exail Quantum Systems, using rubidium atoms cooled by isotropic laser light. MuClock is comparable with typical hydrogen masers in terms of volume and long-term frequency stability, as it nominally reaches a fractional frequency stability of 1 × 10−15 in less than two days and maintains this level of performance over more than one month of integration time. The evaluation of several systematic frequency shifts is on-going. This article presents the impact of the frequency bias induced by microwave phase transients on the clock frequency stability and accuracy. Using an in-house designed test bench, the phase of the 6.8 GHz microwave clock signal is measured with a phase resolution close to 1 μrad and a temporal resolution below 100 ns. It enables a thorough optimization of the sequence and a better control of each phase transient event, making their contributions to the overall frequency shift negligible. The accuracy of the associated systematic effect is evaluated at 3 × 10−15
Fouilles programmées 2023 sur le site de Balchiria (Sartène, Corse du sud) : un centre cérémoniel de l’horizon pré-mégalithique
No full textNational audienceLa découverte fortuite, en 2016, de deux stèles, dont une présentant une iconographie inédite en Méditerranée, a motivé la mise en place d’une série d’opérations de fouilles programmées sur le site de Balchiria, sur la communede Sartène, en Corse-du-Sud. Les opérations conduites entre 2020 et 2023 sous la codirection de la DRAC de Corse et l’Université de Strasbourg, ont permis de contextualiser ces monuments en mettant en évidence plusieursarchitectures de pierres dont certaines peuvent être identifiées à des podiums. Le plus spectaculaire d’entre-deux, de plan naviforme, a conservé une douzaine de petits monolithes découverts en place, encore dressés ou légèrement inclinés. Deux aires empierrées voisines, de plan circulaire ou quadrangulaire sont, en toute hypothèse, assimilées soit à des podiums, soit à des aires cérémonielles
Coupled Hydraulic-Geochemical Processes in a Faulted Clay-Rich Caprock: Reactive Transport Modelling of an In Situ CO2 Injection Experiment
No full textInternational audienceClay–rich formations are widely regarded as effective caprocks for geological CO2 storage; however, the presence of fractures in fault zone introduces significant uncertainty regarding their hydraulic behaviour during injection. In such settings, fluid migration is expected to be largely governed by fracture networks, while coupled hydraulic and geochemical processes associated with CO2–water–rock interactions may progressively modify hydraulic properties over time. This study presents a reactive transport modelling investigation of the CS–D (Carbon Sequestration – Series D) in situ CO2 injection experiment, with the aim of quantifying fracture hydraulic behaviour within a faulted clay–rich caprock.A three-dimensional reactive transport model has been fully implemented using HYTEC to simulate coupled fluid flow and CO2–water–rock interactions within a fractured fault zone embedded in a low-permeability clay matrix. The modelling framework accounts for aqueous speciation, mineral dissolution and precipitation, as well as advective–diffusive transport, and is configured to reproduce the experimental conditions of the CS–D test. Structural and hydrogeochemical observations derived from the experiment are used to constrain boundary conditions and initial states, while fracture hydraulic properties are treated as key uncertain parameters.The numerical framework enables a systematic investigation of the sensitivity of pressure evolution and geochemical responses to variations in fracture permeability and reactive surface area. The current geometric representation captures the principal structural characteristics of the faulted zone and retains sufficient flexibility to explore alternative conceptual configurations as the analysis progresses. The present work addresses coupled hydraulic and geochemical processes and is intended to serve as a basis for future extensions towards a coupled thermo-hydro-mechanical-chemical (THMC) framework relevant to subsurface energy applications.The resulting simulations are expected to provide quantitative constraints on the range of fracture hydraulic properties compatible with the hydraulic and geochemical signals observed during CO2 injection. Ultimately, this study seeks to improve the process-based understanding of fracture-controlled flow in faulted clay-rich caprocks and to support the interpretation of in situ experiments relevant to the long-term integrity and safety of geological CO2 storage and related geo-energy technologies
Intelligence artificielle et production des connaissances scientifiques. Synthèse de la 10e journée nationale d'étude des Urfist (2025)
No full textSynthèse de la journée nationale d'étude des Urfist du 25/11/2025.Depuis le lancement de ChatGPT et des bouleversements que l’intelligence artificielle (IA) générative provoque dans le champ scientifique, les collectifs de recherche et les réseaux professionnels s’emparent du sujet. C’est dans la continuité des thématiques de ses précédentes journées nationales d'étude que le réseau des Urfist a voulu mettre en lumière quelques-uns des défis et opportunités que l’IA générative présente pour la production et la diffusion des connaissances scientifiques. Articulant différents regards, analyses et retours d’expérience, la 10e journée nationale d'étude des Urfist qui s'est tenue le 25 novembre 2025 sur le Campus Condorcet a été introduite par une conférence d’ouverture et terminée par une carte blanche et une conclusion. Trois axes, proposés dans trois sessions différentes, se sont entrecroisés au fil des interventions : celui des processus de production de la recherche, envisageant jusqu’à quel point les IA génératives transforment les opérations de réflexion, de création, de nouveauté et de découverte scientifiques ; celui des processus de diffusion scientifique, avec l'étude de leurs effets dans la validation et la communication de la recherche ; celui enfin des enjeux institutionnels et politiques et du cadre possible pour des usages scientifiques
Diversity comes at a cost: multifaceted diversity reduces plant community stability in peatlands
No full text1.Aim: Understanding how ecological stability relates to diversity is of crucial importance under global change. Greater biodiversity is expected to stabilize aggregate community properties through compensatory dynamics, yet diversity-stability relationships can vary across ecosystems, particularly in wetlands where strong abiotic filters shape community assembly and temporal dynamics. We examined how multiple facets of diversity (taxonomic, functional, and phylogenetic) and functional trait identity relate to temporal stability (invariability) and species asynchrony in peatland vegetation.2.Location: Forbonnet peatland, Jura Mountains, northeastern France. 3.Methods: We used a 17-year field experiment in a montane peatland complex spanning a bog and a transitional poor fen, combining passive open-top chamber warming with natural hydrological contrasts.4.Results: Water table depth was the dominant environmental filter of plant communities, explaining 46 % of total compositional variance, whereas experimental warming had no detectable effect. Community temporal stability and species asynchrony were higher under drier conditions (deeper water table), consistent with moisture-driven constraints on peatland vegetation dynamics. Contrary to insurance hypothesis predictions, temporal stability decreased with multiple biodiversity facets, particularly phylogenetic diversity and species richness, but increased with deeper-rooted plant strategies and functional redundancy, after controlling for experimental conditions. Species asynchrony was largely unrelated to biodiversity, except for functional redundancy, which was negatively associated with asynchrony. The stability-asynchrony association weakened substantially after controlling for hydrology.5.Conclusions: Our results suggest that in peatlands, hydrology simultaneously structures biodiversity patterns, temporal stability and species asynchrony, yielding negative diversity-stability relationships that contradict classical insurance hypothesis predictions. Stability arises primarily from hydrological constraints, with limited contribution from compensatory dynamics among plant species. In strongly constrained, species-poor ecosystems, conservation may therefore prioritize maintaining or restoring the key abiotic conditions that favor functionally adapted communities over increasing diversity to sustain stable ecosystem functioning under global change
Kernel ridge regression under power-law data: spectrum and generalization
No full textIn this work, we investigate high-dimensional kernel ridge regression (KRR) on i.i.d. Gaussian data with anisotropic power-law covariance. This setting differs fundamentally from the classical source & capacity conditions for KRR, where power-law assumptions are typically imposed on the kernel eigenspectrum itself. Our contributions are twofold. First, we derive an explicit characterization of the kernel spectrum for polynomial inner-product kernels, giving a precise description of how the kernel eigenspectrum inherits the data decay. Second, we provide an asymptotic analysis of the excess risk in the high-dimensional regime for a particular kernel with this spectral behavior, showing that the sample complexity is governed by the effective dimension of the data rather than the ambient dimension. These results establish a fundamental advantage of learning with power-law anisotropic data over isotropic data. To our knowledge, this is the first rigorous treatment of non-linear KRR under power-law data
Le Capital d'une capitale
No full textInternational audienceA travers l'exemple de Paris, ce livre propose une lecture renouvelée de l'histoire de la richesse. Grâce à une documentation exceptionnelle, l'analyse de l'évolution de la richesse parisienne au cours des deux derniers siècles permet en effet d'éclairer des processus rarement identifiables ailleurs. Dans la capitale, autant l'importance et la composition de la richesse varient, autant sa distribution reste stable. Les données individuelles portant sur l'ensemble de la population montrent que la richesse et ceux qui la détiennent connaissent un renouvellement continu, y compris au sein des plus riches, alors même qu'une part toujours considérable de la richesse est héritée. Si les Parisiens sont nombreux à épargner, leurs efforts sont épuisés par les incertitudes de la vie de sorte que, jusque dans les années 1920, deux tiers d'entre eux n'ont aucune richesse au décès. Il faut attendre la diffusion des systèmes de retraite pour que cette proportion diminue. Seule l'intervention publique permet alors de réduire la part très élevée des personnes qui meurent sans richesse ainsi que les écarts d'espérance de vie, de santé et d'éducation. Ces dimensions essentielles, longtemps presque figées en dépit de la longue croissance de la richesse jusqu'à la Première Guerre mondiale, n'ont sensiblement progressé qu'ensuite, notamment lorsque l'Etat social a transformé les conditions d'accumulation et de transmission du capital. L'histoire de la richesse ouvre ainsi sur une histoire des choix politiques, des aspirations sociales et des inégalités que les sociétés acceptent - ou refusent - de tolérer
Quand le genre fait parole. Pour une archéologie de l'éducation féminine antique.
No full textInternational audienceThis communication focuses on the understanding of ancient female education, in fifth and fourth centuries BCE in Greece, through the lens of gender. Ancient female education did exist and deserves to be rendered visible through the examination of diverse archaeological artefacts. This study sheds light on ancient female education by following a progression structured around three stages in women’s lives: first, the gendered distinctions of the παῖς (pais), the child; second, the pedagogical practices associated with the παρθένος (parthenos), the unmarried young woman; and finally, the γυνή (gunê), the woman as educator. Approaching ancient female education through archaeology and a gendered perspective enables us to restore a voice to those who have remained inaudible within ancient literary history.Cette communication s’intéresse à l'appréhension de l'éducation féminine antique, dans la Grèce des Ve et IVe siècles a.C., au prisme du genre. L'éducation féminine antique existe et mérite d'être rendue visible par l'appréhension de différents vestiges archéologiques. Cette communication donne à voir l'éducation féminine antique en suivant une progression fondée sur trois périodes de la vie des femmes : dans un premier temps autour des distinctions genrées du παῖς (pais) de l'enfant, dans un second temps par les pratiques pédagogiques associées à la παρθένος (parthenos) jeune fille non mariée et, enfin, dans un dernier temps par la γυνή (gunê) femme éducatrice. S'intéresser à l'éducation féminine antique, par l'archéologie et au prisme du genre, permet de redonner voix à des inaudibles de l'histoire littéraire ancienne