HAL-MINES ParisTech
Not a member yet
27348 research outputs found
Sort by
Spline Interpolation on Compact Riemannian Manifolds
No full textSpline interpolation is a widely used class of methods for solving interpolation problems by constructing smooth interpolants that minimize a regularized energy functional involving the Laplacian operator. While many existing approaches focus on Euclidean domains or the sphere, relying on the spectral properties of the Laplacian, this work introduces a method for spline interpolation on general manifolds by exploiting its equivalence with kriging. Specifically, the proposed approach uses finite element approximations of random fields defined over the manifold, based on Gaussian Markov Random Fields and a discretization of the Laplace-Beltrami operator on a triangulated mesh. This framework enables the modeling of spatial fields with local anisotropies through domain deformation. The method is first validated on the sphere using both analytical test cases and a pollution-related study, and is compared to the classical spherical harmonics-based method. Additional experiments on the surface of a cylinder further illustrate the generality of the approach
Courants de Gravité : Mesures combinées de densité et de vélocité pour le calcul du champ de flottabilité
No full textInternational audienceThe understanding of gravity currents is required for a proper characterization of several natural andindustrial phenomena. They induce a strong mixing within turbulent flows that are stratified in temperature, solute or suspended particles. This mixing is intrinsically complex and requires in-depth experimental studies. In particular, although we can acquire the dynamic of both density and velocity fields,only simulations allowed us to visualize these fields simultanuously to understand their interaction. Thatis why we thereby propose an acquisition method to get the density and velocity fields on a single flowto deduce, for instance, the dynamics of buoyancy fields.La compréhension des courants de gravité est nécessaire à la bonne caractérisation d’un grand nombrede phénomènes naturels et industriels. Ils induisent un fort mélange au sein des écoulements turbulents stratifiés en température, en présence de soluté ou en suspension de particules. Ce mélange est intrinsèquement complexe et requiert des études expérimentales approfondies. En particulier, bien qu’onconnaisse respectivement la dynamique des champs de densité et de vélocité, seules les simulations nouspermettaient de visualiser ces champs simultanément pour appréhender leur interaction. C’est pourquoinous proposons ici une méthode pour acquérir les champs de densité et de vélocité sur un même écoulement pour en déduire entre autres, la dynamique des champs de flottabilité
Prospective Life Cycle Assessment of French electricity scenarios by 2050
No full textInternational audienceWhat is the influence of imports and storage for Global Warming Potential of French power consumption mix for scenarios with high shares of renewables in 2050 ? • The environmental assessment is made in the framework of prospective Life Cycle Assessment• Neglecting imports and storage can lead to underestimating the GWP of prospective national consumption mix with high share of fluctuating renewables.• However even when including storage and import, all these scenarios have much lower GWP per kWh than current electricity mixes (FR & EUR)
NICE k metrics: Unified and multidimensional framework for evaluating deterministic solar forecasting accuracy
No full textInternational audienc
Improving Morphological Networks for Learning Image-to-Image Transforms
No full textInternational audienceReplacing convolution with morphological operations in trainable layers has received significant attention lately. Among the various strategies that have emerged, smooth morphological layers have shown strong potential and flexibility, as a single layer can behave either like a (pseudo-)erosion or a (pseudo-)dilation depending on the sign and value of its trainable control parameter. In this work, we build upon the so-called SMorph layer by introducing a harmonized formulation that addresses previously identified asymptotic limitations when learning grayscale erosion and dilation. We also investigate and compare two strategies (a novel penalty term in the training loss and shared-weight layers) to improve the learning of grayscale opening and closing operations in two-layer networks. Finally, we evaluate the performance of this improved SMorph layer on a salt-and-pepper denoising task in a four-layer network architecture, and compare it with other morphological and convolutional networks
One Wave to Explain Them All: A Unifying Perspective on Feature Attribution
No full textInternational audienceFeature attribution methods aim to improve the transparency of deep neural networks by identifying the input features that influence a model's decision. Pixel-based heatmaps have become the standard for attributing features to high-dimensional inputs, such as images, audio representations, and volumes. While intuitive and convenient, these pixel-based attributions fail to capture the underlying structure of the data. Moreover, the choice of domain for computing attributions has often been overlooked. This work demonstrates that the wavelet domain allows for informative and meaningful attributions. It handles any input dimension and offers a unified approach to feature attribution. Our method, the Wavelet Attribution Method (WAM), leverages the spatial and scale-localized properties of wavelet coefficients to provide explanations that capture both the where and what of a model's decision-making process. We show that WAM quantitatively matches or outperforms existing gradient-based methods across multiple modalities, including audio, images, and volumes. Additionally, we discuss how WAM bridges attribution with broader aspects of model robustness and transparency. Project page: https://gabrielkasmi.github.io/wam
EU Digital Technologies and Policy Conference (EUDTP 2025) Abstracts and Contributions
No full textInternational audienc
A Deep Learning approach for time-consistent cell cycle phase prediction from microscopy data
No full textAbstract The cell cycle consists of four phases and impacts most cellular processes. In imaging assays, the cycle phase can be identified using dedicated cell-cycle markers. However, such markers occupy fluorescent channels that may be needed for other reporters. Here, we propose to address this limitation by inferring the phase from a widely used fluorescent reporter: SiR-DNA. Our method is based on a variational auto-encoder, enhanced with two auxiliary tasks: predicting the intensity of phase-specific markers and enforcing the latent space temporal consistency. Our model is freely available, along with a new dataset comprising over 600,000 annotated HeLa Kyoto nuclear images
Galaxy–point spread function correlations as a probe of weak-lensing systematics with UNIONS data
No full textInternational audienceContext. Weak gravitational lensing requires precise measurements of galaxy shapes and therefore accurate knowledge of the point spread function (PSF) model. The latter can be a source of systematics that affect the shear two-point correlation function. A key aspect of weak-lensing analysis is the forecasting of the systematics due to the PSF.Aims. Correlation functions of galaxies and the PSF, the so-called ρ and τ statistics, are used to evaluate the level of systematics coming from the PSF model and PSF corrections and contributing to the two-point correlation function used to perform cosmological inference. Our goal is to introduce a fast and simple method to estimate this level of systematics and to assess its agreement with state-of-the-art approaches.Methods. We introduce a new way to estimate the covariance matrix of τ statistics using analytical expressions. The covariance allows us to estimate parameters directly related to the level of systematics associated with the PSF and provides us with a tool to validate the PSF model used in a weak-lensing analysis. We applied these methods to data from the Ultraviolet Near-Infrared Optical Northern Survey (UNIONS).Results. We show that semi-analytical covariance yields results comparable to those obtained by using covariances obtained from simulations or jackknife resampling. The approach requires less computation time and is therefore well suited to rapid comparison of the systematic level obtained from different catalogues. We also show how one can break degeneracies between parameters with a redefinition of the τ statistics.Conclusions. The methods developed in this work will be useful tools in the analysis of current weak-lensing data but also of Stage IV surveys such as Euclid, LSST, and Roman. They provide fast and accurate diagnostics on PSF systematics that are crucial in the context of cosmic shear studies