HAL UVSQ
Not a member yet
57319 research outputs found
Sort by
ArchesWeather & ArchesWeatherGen: a deterministic and generative model for efficient ML weather forecasting
No full textInternational audienceWeather forecasting plays a vital role in today's society, from agriculture and logistics to predicting the output of renewable energies, and preparing for extreme weather events. Deep learning weather forecasting models trained with the next state prediction objective on ERA5 have shown great success compared to numerical global circulation models. However, for a wide range of applications, being able to provide representative samples from the distribution of possible future weather states is critical. In this paper, we propose a methodology to leverage deterministic weather models in the design of probabilistic weather models, leading to improved performance and reduced computing costs. We first introduce \textbf{ArchesWeather}, a transformer-based deterministic model that improves upon Pangu-Weather by removing overrestrictive inductive priors. We then design a probabilistic weather model called \textbf{ArchesWeatherGen} based on flow matching, a modern variant of diffusion models, that is trained to project ArchesWeather's predictions to the distribution of ERA5 weather states. ArchesWeatherGen is a true stochastic emulator of ERA5 and surpasses IFS ENS and NeuralGCM on all WeatherBench headline variables (except for NeuralGCM's geopotential). Our work also aims to democratize the use of deterministic and generative machine learning models in weather forecasting research, with academic computing resources. All models are trained at 1.5° resolution, with a training budget of 9 V100 days for ArchesWeather and 45 V100 days for ArchesWeatherGen. For inference, ArchesWeatherGen generates 15-day weather trajectories at a rate of 1 minute per ensemble member on a A100 GPU card. To make our work fully reproducible, our code and models are open source, including the complete pipeline for data preparation, training, and evaluation, at https://github.com/INRIA/geoarches
Unveiling the Power of Early Preschool Education: A Transformative Case Study from Vietnam
No full textEarly childhood, known as the cornerstone of human development, stands to gain unprecedented benefits from effective preschool education. In the context of Vietnam, a country witnessing rapid economic and social changes, the timely commencement of preschool education is nothing short of a revolution in nurturing young minds. Our study meticulously delves into the paradigm shift brought about by early preschool education, furnishing compelling evidence that underscores its unparalleled potential to foster cognitive, social, and emotional growth in the developmental trajectory of children. By exploring this transformative case of Vietnam, we aim to illuminate the quintessential role of preschool education in creating a generation of well-rounded individuals poised to thrive in an increasingly complex world.</div
Canadian net forest CO2 uptake enhanced by heat drought via reduced respiration
No full textInternational audienc
Microbial life in Arctic pack ice: Prospects for the Tara Polaris expeditions
No full textInternational audienceSea ice, unlike freshwater ice, hosts abundant microbial life, thanks to the presence of liquid water inclusions encased within the ice matrix. The forthcoming Tara Polaris Expeditions (TPE), which will document drifting Arctic pack ice repeatedly over multiple years, together offer an unprecedented opportunity to advance understanding of the sea-ice microbiome -its diversity, variations, and ecological roles. In this contribution, we consider the current state of knowledge, identify key research gaps, and outline the potential for progress enabled by TPE. We envision the emergence of new insights into the seasonal evolution of microbial life, resolved at the floe (kilometric) scale, in relation to the evolution of the sea-ice environment -its morphology, light, temperature, and liquid water distribution and properties. Large potential lies in the characterization of diverse microhabitats across the central Arctic Ocean, associated with brine inclusions, pressure ridge cavities, and melt ponds. A major goal will be to document biological processes that remain poorly understood -colonization, diversity, functioning, interactions, and evolutionary dynamics -and that could benefit from the application of newly developed techniques. We argue that TPE is particularly timely, as the loss of multi-year ice may soon constrain opportunities to study life in this rapidly changing habitat
Toward dimensional body consciousness impairments in post-traumatic stress disorder and its dissociative subtype: a predictive processing approach
No full textInternational audienceCurrent models of post-traumatic stress disorder (PTSD) predominantly focus on emotional dysregulation, avoidance behaviors, and intrusive symptoms, with limited attention to disturbances in bodily self-consciousness. These frameworks do not fully account for how trauma may disrupt the integration of interoceptive and exteroceptive signals underlying the sense of embodied self. Building on recent advances in computational psychiatry, this review introduces a novel theoretical framework that differentiates between dissociative and non-dissociative forms of PTSD using the predictive processing paradigm. Body awareness, including the sense of body ownership (SBO) and the sense of agency (SoA), is framed within a Bayesian inference framework, where perception results from the interaction between prior beliefs and incoming sensory information. In non-dissociative PTSD, we suggest that hyperprecise trauma-related priors, coupled with increased interoceptive weighting due to amygdala and anterior insula hyperactivity, result in rigid self-representations and a diminished capacity for perceptual updating. Thus, we may consider PTSD as a state in which accuracy and reliability of cognitive processes are ranked as follows: [Prior > Interoception > Exteroception]. Conversely, in the dissociative PTSD subtype, emotional over-inhibition and anterior insula hypoactivity weaken priors and interoception, while exteroceptive inputs dominate. Therefore, we consider ranked cognitive processes in PTSD dissociative subtype as follows: [Exteroception > Interoception > Prior]. SoA impairements is specific to the dissociative subtype involving hyperactivity of the angular gyrus and glutamate hypofunction. On this basis, we propose an original dimensional model of body consciousness disruption across PTSD spectra. Therapeutic implications are explored, including top-down and bottom-up interventions
Volcanoes: Composition of Emissions
No full textInternational audienceVolcanic emissions mainly consist of water vapor, carbon dioxide, and sulfur gases, with smaller amounts of halogens and other trace gases. Their composition varies depending on the magma’s source and its storage conditions in the crust. Large explosive eruptions can release gases directly into the stratosphere, affecting atmospheric composition, especially the ozone layer, and altering the global climate for years. In recent decades, most global volcanic impacts have been caused by stratospheric aerosols formed from sulfur gas oxidation, though water vapor and halogens may also contribute to atmospheric changes
Meta-Analysis and Bioinformatic Analysis of Human Fluid Proteomic Studies Reveal Key Biological Pathways Related to Aging and Frailty
No full textInternational audienceAging research seeks to understand the progressive decline in physiological function throughout adulthood. Frailty is a significant clinical manifestation of aging. Recent discoveries of global hallmarks of aging have paved the way for deeper insight into its molecular mechanisms. Proteomics has emerged as a powerful approach, given the proteome's sensitivity to environmental changes. This study aimed to uncover novel biological insights into aging by analyzing proteins that are significantly associated with chronological age or frailty. We conducted a systematic review of Medline and Embase up to August 2024 to find studies involving adults that used proteomic analyses. From each study, we compiled the significant proteins consistently reported across age-related or frailty-related studies by extracting statistical outcomes and expression patterns. Significant proteins were compared, and pathway enrichment analysis was performed using STRING software. Our analysis included 2,630 age-associated proteins from 27 studies and 194 frailty-associated proteins from 8 studies. We identified 177 shared proteins for both age and frailty, including nine with opposite expression trends. Meta-regression and pathway analyses revealed convergence on key biological processes, including immunity, inflammation, metabolism, homeostasis, coagulation, and neurology. These findings suggest that age- and frailty-related proteomic studies provide complementary insights into understanding the overall complexity of the biology of aging. This integrative approach emphasizes the potential of proteomics not only for biomarker discovery but also for advancing our understanding of our functional capacities. Future multi-omics studies will be essential for further elucidating the complex molecular landscape of aging
Co-design and evaluation with construction workers of virtual reality and force feedback modalities for formwork panel action training
No full textInternational audienceAbstract Construction is one of the sectors most at risk in terms of health and safety. The social security system attributes more than 14% of the accidents recorded in 2018 to the construction sector. In addition to safety risks, construction workers are exposed to occupational diseases. Musculoskeletal disorders (MSD) are at the top list of the most frequent occupational diseases. In order to address these issues, construction companies have implemented health and safety policy to protect their employees. This policy provides, among other things, for an improvement in the organizational aspect of work, the adaptation of work to people and training. Training is carried out according to a classic methodology: theoretical and practical training. The training represents constraints of cost, duration and organization. The content must be adapted to specific profiles, be repeatable and allow the trainers to assess the level of the trainees. For training on the right actions and posture to avoid the risk of MSDs, the impact of virtual reality training is less relevant due to the lack of haptic feedback. In this article, we propose to evaluate with the construction workers a new training methodology based on visual immersion and a robotic platform for haptic interaction with motion tracking. This methodology is evaluated for a particular task: the formwork panel adjustment, a highly risky operation on a highly sensitive tool. We believe that this approach would allow to better engage the construction workers during training on the formwork, to ensure their safety and awareness of MSD risks. The objective through this method is to enable the construction workers to understand the risks linked to their job, to train them to avoid risks, and to enable the trainer to benefit from a tool for monitoring and supporting the construction workers. In order to validate the methodology, experiments were conducted and questionnaires were submitted to two profiles of people: novices and professionals. This experimentation shows an acceptance of the solution by the construction workers
Possible favored great oxidation event scenario on exoplanets around M-stars with the example of TRAPPIST-1e
No full textInternational audienceThe Great Oxidation Event (GOE), which marked the transition from an anoxic to an oxygenated atmosphere, occurred 2.4 billion years ago on Earth, several hundreds of millions of years after the emergence of oxygenic photosynthesis. This long delay implies that specific conditions in terms of biomass productivity and burial were necessary to trigger the GOE. It could be a limiting factor for the development of oxygenated atmospheres on inhabited exoplanets. In this study, we explore the specificities of a terrestrial planet in the habitable zone of an M dwarf for a GOE. Using a 1D coupled photochemical-climate model, we simulate the atmospheric evolution of TRAPPIST-1e, an Earth-like exoplanet, exploring the effect of oxygen sources (biotic or abiotic). Our results show that the stellar energy distribution promotes O[Formula: see text] production at lower O[Formula: see text] concentrations compared to Earth, and the ozone layer on TRAPPIST-1e forms more efficiently. This lowers the threshold for atmospheric oxidation, suggesting that the GOE on TRAPPIST-1e would occur quickly after the rise of oxygenic photosynthesis, up to 1Gyrs earlier than on Earth, and would reach O[Formula: see text] enabling oxygenic respiration and thus the development of animals. We may question whether this is a general behavior around several M-stars. Furthermore, we discuss how the overproduction of ozone could make O[Formula: see text] detection possible using the James Webb Space Telescope, providing a potential method to observe oxygenation signatures on exoplanets in the near future. Previous studies predicted that for an Earth-like atmosphere O[Formula: see text] would require over 150 transits for detection, but our results show that significantly fewer transits could be needed
