HAL-Ecole des Ponts ParisTech
Not a member yet
42743 research outputs found
Sort by
Data Science: a Natural Ecosystem
No full textInternational audienceThis manuscript provides a systemic and data-centric view of what we term essential data science, as a natural ecosystem with challenges and missions stemming from the fusion of data universe with its multiple combinations of the 5D complexities (data structure, domain, cardinality, causality, and ethics) with the phases of the data life cycle. Data agents perform tasks driven by specific goals. The data scientist is an abstract entity that comes from the logical organization of data agents with their actions. Data scientists face challenges that are defined according to the missions. We define specific discipline-induced data science, which in turn allows for the definition of pan-data science, a natural ecosystem that integrates specific disciplines with the essential data science.We semantically split the essential data science into computational, and foundational.By formalizing this ecosystemic view, we contribute a general-purpose, fusion-oriented architecture for integrating heterogeneous knowledge, agents, and workflows—relevant to a wide range of disciplines and high-impact applications
Advanced insights into the biodeterioration and conservation strategies of cultural heritage: A review
No full textInternational audienceStone monuments, which have immense historical, artistic, and cultural significance, are vulnerable to environmental factors and microbial colonization, leading to irreversible material loss and posing major challenges for cultural heritage conservation worldwide. Microbial colonization of stone monuments is highly influenced by environmental factors causing subsequently stone degradation through acid production, mineral dissolution, and mechanical stress. Despite numerous studies, a critical synthesis linking microbial mechanisms with conservation strategies remains limited. This review provides a comprehensive analysis of both traditional and modern conservation techniques to prevent, limit or repair the negative effects caused by microorganisms. Conventional biocidal treatments, though effective initially, raise concerns regarding long-term performance and ecological safety, prompting research into sustainable alternatives such as nanoparticles, polymer-based coatings, and natural antimicrobial agents. Emerging approaches, including the use of nanoparticles, polymer-based protective coatings, and natural antimicrobial agents such as essential oils, are reviewed in detail. Special emphasis focuses on evaluating the efficacy, durability, and environmental impact of these interventions. This review highlights the need for eco-compatible conservation solutions that balance antimicrobial efficiency with material durability and environmental safety
19F magnetic resonance imaging-informed fate models of PFAS in porous media
No full textInternational audiencePer-and polyfluoroalkyl substances (PFAS) are persistent contaminants. Predicting their fate in natural or engineered porous media, using accurate models, is essential for effective remediation and contamination management strategies. The mechanisms of transport and retention included in such models, and the associated parameters, are mostly inferred from PFAS concentration vs. time breakthrough curves (BTCs) measured during transport experiments. Still, the interpretation of BTCs may not be unique as they result from a succession of mechanisms taking place inside the porous media. We addressed this issue using 19 F magnetic resonance imaging (MRI) to monitor the transport of perfluorobutanoic acid (PFBA) inside a sand-packed column. The experimental BTC was slightly asymmetric, suggesting that some PFBA may have been adsorbed onto the sand. Hence, a transport model based on the hypothesis that PFBA behaved as a non-sorbing tracer slightly overestimated the concentrations in two regions of the BTC. Surprisingly, the same model matched well the MRI profiles, pointing out that the BTC asymmetry stemmed from an imperfect column exit. Although 19 F MRI requires PFAS concentration above those found in environmental samples, this study showed that the combination of this technique and modeling constitutes a powerful tool to determine the mechanisms involved in PFAS transport in natural or engineered porous media andselect appropriate fate model
Explaning with trees: interpreting CNNs using hierarchies
No full textInternational audienceChallenges persist in providing interpretable explanations for neural network reasoning in explainable AI (xAI). Existing methods like Integrated Gradients produce noisy maps, and LIME, while intuitive, may deviate from the model’s reasoning. We introduce a framework that uses hierarchical segmentation techniques for faithful and interpretable explanations of Convolutional Neural Networks (CNNs). Our method constructs model-based hierarchical segmentations that maintain the model’s reasoning fidelity and allows both human-centric and model-centric segmentation. This approach offers multiscale explanations, aiding bias identification and enhancing understanding of neural network decision-making. Experiments show that our framework, xAiTrees, delivers highly interpretable and faithful model explanations, not only surpassing traditional xAI methods but shedding new light on a novel approach to enhancing xAI interpretability
Trade agreements and sustainable fisheries
No full textInternational audienceThis study examines the impact of trade agreements and their specific provisions on the sustainability of marine fisheries resources. Using global data on the Mean Trophic Level (MTL) between 1950 and 2018 and a comprehensive dataset of environmental provisions from trade agreements signed between 1947 and 2018, we estimate the impact on the MTL of signing (i) a free trade agreement and (ii) a free trade agreement including fishery-related provisions. To address potential endogeneity problems associated with fisheries-related provisions, we use a difference-in-differences (DID) propensity score matching method. Our results show that while trade agreements tend to negatively impact the MTL, including fisheries-related provisions offsets this negative impact among signatory countries. By examining the potential mechanisms underlying this result, we are able to temper the optimistic findings in the existing literature on the beneficial environmental outcomes of environmental provisions. Our findings suggest that these provisions do not foster the adoption of more effective resource management practices. Instead, they appear to reduce trade opportunities, which is contrary to the objective of trade creation in trade agreements
Coarse-to-fine crack cue for robust crack detection
No full textInternational audienceCrack detection is an important task in computer vision. Despite impressive in-dataset performance, deep learning-based methods still struggle in generalizing to unseen domains. The thin structure property of cracks is usually overlooked by previous methods. In this work, we introduce CrackCue, a novel method for robust crack detection based on coarse-to-fine crack cue generation. The core concept lies on leveraging the thin structure property to generate a robust crack cue, guiding the crack detection. Specifically, we first employ a simple max-pooling and upsampling operation on the crack image. This results in a coarse crack-free background, based on which a fine crack-free background can be obtained via a reconstruction network. The difference between the original image and fine crack-free background provides a fine crack cue. This fine cue embeds robust crack prior information which is unaffected by complex backgrounds, shadow, and varied lighting. As a plug-and-play method, we incorporate the proposed CrackCue into three advanced crack detection networks. Extensive experimental results demonstrate that the proposed CrackCue significantly improves the generalization ability and robustness of the baseline methods. The source code is publicly available at https://github.com/LZL501/CrackCue
Simultaneous estimation of radiance and its sensitivities to radiative properties in a spherical-heterogeneous atmospheric radiative transfer model by Monte Carlo method: Application to Titan
No full textInternational audienceWe propose a control variates technique to reduce the variance of null-collision Monte Carlo algorithms used for solving the Radiative Transfer Equation (RTE) in highly heterogeneous media. The method complements the classical spatially partitioned overestimate approach by additionally recording the minimum absorption coefficient within each voxel during preprocessing. During path tracing, the attenuation due to this minimum absorption is evaluated analytically, while the residual part is handled by path-samplings. This analytical treatment significantly improves convergence particularly in strongly absorbing media such as the planetary atmospheres in infrared absorbing band. The mathematical equivalence between the original and control-variates estimators is demonstrated, and numerical applications for Earth's and Titan's atmospheres confirm the expected variance reduction.</div
A Soil–Plant–Atmosphere Continuum model coupled to CFD to simulate plant energy and water exchanges in heterogeneous microclimates
No full textInternational audienceEstimating plant growth conditions in agrivoltaic, agroforestry, or urban environments are applied examples exhibiting the need to consider the intricate relationships between spatially heterogeneous microclimate conditions (short-wave and long-wave radiation, wind, turbulence, and air temperature), plant and soil energy balances with air and water exchanges. To capture these connections, the Soil–Plant–Atmosphere Continuum model from A. Tuzet has been implemented in the computational fluid dynamics software code_saturne, which simulates spatially heterogeneous and time-varying fluid flows, along with short-wave and long-wave radiation. This coupling is compared to experimental measurements from two French sites of the Integrated Carbon Observatory System (ICOS). Our model achieves significant outcomes in assessing energy exchanges, maintaining a relative error of less than 20% compared to ICOS measurements. In addition to accurately reproducing variations of latent and sensible heat fluxes due to radiation, the coupling of the water balance and stomatal conductance models demonstrates its capability to predict the evolution of soil water content over several days. Finally, an extrapolative study of fictive environments with plants beneath obstacles reveals promising opportunities to understand how obstacle-induced shadows and wakes affect plant temperature. This leads the way for further research in agrivoltaic, agroforestry, or urban configurations with spatial scales from approximatively 10m2 up to 1000m2 and temporal scales ranging from single moments to several consecutive days
Explaining is not enough: Appealing explanations should also be surprising
No full textInternational audiencePhilosophers have attempted to define the features that make an explanation a good explanation, and psychologists have shown that people are sensitive to many of these features.Psychologists have also pointed out the importance of the phenomenology of explanations: the pleasure we derive from formulating or encountering good explanations would motivate us to seek more explanations. However, it seems that many good explanations do not trigger such positive feelings: they are good explanations, but they are not particularly appealing. We suggest that for an explanation to be appealing, it should not only explain the relevant phenomenon (be explanatory), but it should also be surprising. This is what we observe in three experiments, using both explanations from past studies, and more ecologically valid explanations gathered on the subreddit Explain Like I'm 5. We also find that the usefulness of the phenomenon being explained is another predictor of the appeal of the explanation. Finally, we show that surprisingness ratings do not depend only on whether the explanation was already known, and that their effect on appeal does not decrease when controlling for prior knowledge.Instead, explanations are judged more surprising when others do not know them, and we hypothesize that internal properties of explanations also play a role.</p
Exploiting Low Scanwidth to Resolve Soft Polytomies
No full textInternational audiencePhylogenetic networks allow modeling reticulate evolution, capturing events such as hybridization and horizontal gene transfer. A fundamental computational problem in this context is the Tree Containment problem, which asks whether a given phylogenetic network is compatible with a given phylogenetic tree. However, the classical statement of the problem is not robust to poorly supported branches in biological data, possibly leading to false negatives. In an effort to address this, a relaxed version that accounts for uncertainty, called Soft Tree Containment, has been introduced by Bentert, Malík, and Weller [SWAT’18]. We present an algorithm that solves Soft Tree Containment in 2^O(∆(T) k log(k)) · n^O(1) time, where k := sw(Γ ) + ∆(N) , with ∆(T) and ∆(N) denoting the maximum out-degrees in the tree and the network, respectively, and sw(Γ ) denoting the “scanwidth” [Berry, Scornavacca, and Weller, SOFSEM’20] of a given tree extension of the network, while n is the input size. Our approach leverages the fact that phylogenetic networks encountered in practice often exhibit low scanwidth, making the problem more tractable