HAL-Polytechnique
Not a member yet
51406 research outputs found
Sort by
Estimation of discrete distributions in relative entropy, and the deviations of the missing mass
No full textInternational audienceWe study the problem of estimating a distribution over a finite alphabet from an i.i.d. sample, with accuracy measured in relative entropy (Kullback–Leibler divergence). While optimal bounds on the expected risk are known, high-probability guarantees remain less well-understood. First, we analyze the classical Laplace (add-one) estimator, obtaining matching upper and lower bounds on its performance and establishing its optimality among confidence-independent estimators. We then characterize the minimax-optimal high-probability risk and show that it is achieved by a simple confidence-dependent smoothing technique. Notably, the optimal non-asymptotic risk incurs an additional logarithmic factor compared to the ideal asymptotic rate. Next, motivated by regimes in which the alphabet size exceeds the sample size, we investigate methods that adapt to the sparsity of the underlying distribution. We introduce an estimator using data-dependent smoothing, for which we establish a high-probability risk bound depending on two effective sparsity parameters. As part of our analysis, we also derive a sharp high-probability upper bound on the missing mass
RHITA: a web tool for real-time detection of extreme weather events
No full textExtreme weather hazards are increasing and stakeholders need rapid, transparent information during unfolding events. We present RHITA (Real-time Hazard Identification and Tracking Algorithm), an open-source framework and web tool for near real-time detection and tracking of weather-related hazards over Europe. RHITA identifies grid cells exceeding local quantile thresholds, groups them into spatial clusters, and links clusters through time to reconstruct three-dimensional events in longitude, latitude, and time. For each event, RHITA provides intensity, extent and duration metrics and estimates rarity through return periods derived from a long historical record.RHITA is operated with ECMWF open forecasts for daily monitoring and ERA5 reanalysis for a consistent historical archive from 1950 to 2024. We target four hazards: heatwaves, cold spells, heavy precipitation and strong winds. Key spatial and temporal parameters are optimized against EM-DAT disaster records (2000 to 2023). Applying RHITA to ERA5 yields a European climatology of hazard events and reveals robust increases in heatwave frequency, intensity and affected area, a decline in cold spell frequency, and more heterogeneous signals for heavy precipitation and strong winds at the continental scale. RHITA provides open access data and an interactive interface to support rapid hazard characterization, event contextualization and downstream risk analysis
Consumer Choice between Recommendation Algorithms - Experimental Evidence
No full textInternational audienceRegulators are increasingly concerned with the power of large online platforms to bias consumer recommendations. In light of these concerns, I study experimentally whether giving consumers the choice between several recommendation algorithms can improve consumer welfare. I develop a novel experimental setting in which I first elicit the subjects' preferences and then expose them to choice tasks intermediated by personalized recommendation algorithms. When faced with a costly choice between two algorithms, subjects have a positive willingness-to-pay for better recommendations. However, subjects underestimate the potential gain from the better algorithm: a \$1 increase in estimated gain raises willingness-to-pay by only \$0.15 on average. These findings suggest that giving consumers power over recommendation algorithms to curtail potential abuse is not straightforward. However, it may be a viable business for platforms to offer improved recommendation algorithms to consumers for a fee
A study of the magnetic effects of the Equatorial Electrojet (EEJ) along the East Asian and West African sectors
No full textInternational audienceThis paper presents a comparative analysis of the Equatorial Electrojet (EEJ) effects along the East Asian (140° E) and West African (5° W) sectors using geomagnetic field data recorded in 1993. The data were obtained within the frameworks of the International Equatorial Electrojet Year (IEEY) campaign for West Africa and the Solar-Terrestrial Energy Program (STEP) project for the Asian sector. The analysis was conducted under quiet magnetic conditions (Am < 20 nT) to facilitate the investigation of regular daily variations () induced by the EEJ. Key electrodynamic parameters of the EEJ such as ribbon width, center position, and maximum current intensity were estimated using latitudinal profiles of the H and Z components of the geomagnetic field. The results reveal significant latitudinal and longitudinal variations characterized by a wider EEJ ribbon in Asia (∼714 km) compared to West Africa (∼601 km), and a higher peak current intensity in West Africa (214.3 ± 30.9 A/km) relative to Asia (142.4 ± 31.9 A/km). Also, in 1993, the EEJ centre was located north of the magnetic equator in West Africa and south of it in Asia, indicating a pronounced hemispheric asymmetry. Seasonal variations show higher EEJ intensities during the equinoxes than during the solstices in both longitude sectors. Throughout 1993, the EEJ peak intensity over West Africa consistently surpassed that over East Asia, irrespective of the magnetic season. Furthermore, the frequency of occurrence of the counter-electrojet (CEJ) was higher in West Africa, especially during the morning hours. This study shows that EEJ morphology and strength are modulated by longitude-dependent factors, including ionospheric conductivity, neutral winds, atmospheric tides, and the geomagnetic field
Stress scenarios of cyber loss processes with dependencies
No full textCyber risk has become one of the most critical threats to organizations, characterized by its contagious nature and its dependence on vulnerability disclosures. These features make it particularly challenging for insurers to quantify systemic exposures and design stress tests that assess portfolio resilience under extreme scenarios. In this paper, we develop a model that explicitly incorporates the role of vulnerabilities in cyber attack dynamics and captures contagion and clustering features. Using the general framework of Multivariate Self-Exciting Processes with Dependencies (MSPD), we provide closed-form expressions for the expectation and the variance of a cyber loss process under an externally excited Hawkes model, and we derive closed-form formulas for the surplus under two stress scenarios: (i) an excess of claims in the portfolio and (ii) a massive disclosure of critical vulnerabilities. This novel approach enables trajectory-based stress tests by evaluating the loss process under deterministic scenarios of claims or vulnerability disclosures. By quantifying the impact of these scenarios, our framework offers insurers practical tools for cyber risk stress testing, portfolio management, and regulatory compliance
One- and Two-Photon Polymerization of Solvent- and Filler-Free Aromatic Organic Precursors: Toward 3D-Printed Semiconducting Microstructures
No full textInternational audienceOver the past decades, a great deal of attention has been dedicated toward developing semiconductive polymers (SP), which are now considered as a critical class of photoactive and electroactive materials. However, the processing of many SPs involves the use of solvents, leading to the fabrication of samples with limited shapes, mostly flat 2D thin films. To develop 3D-shaped materials with potential semiconducting properties, this work tackles the development of solvent-and filler-free resins for the fabrication via additive manufacturing of microscale organic semiconductors. Two molecular liquids based on electron-rich carbazole and triarylamine units are used to formulate photoresists. The successful reactivity of the formulations was investigated, both at the macroscopic scale (one-photon polymerization through UV curing) and at the microscopic scale (two-photon polymerization (2PP) via direct laser writing), with Fourier transform infrared spectroscopy (FT-IR) and Raman spectroscopy. Subsequently, the thermal properties of the macroscopic samples were investigated by differential scanning calorimetry (DSC) and thermogravimetric analysis (TGA). Optimization of the printing parameters for 2PP led to the elaboration of 3D micrometer-scale samples whose morphology was assessed by scanning electron microscopy (SEM). Finally, electrical measurements revealed a semiconducting behavior as the samples were found to conduct current after p-doping with iodine
Probabilistic Heat Transfer Coupled For Improved Urban Climate Predictions
No full textOur cities are getting hotter, particularly during heatwaves, which impacts our comfort and energy bills. It is crucial to predict how heat moves through complex urban areas, from buildings to streets, in order to design cooler, more sustainable cities. However, current methods struggle to cope with the complexity of urban landscapes. This paper introduces a revolutionary computer model that addresses this challenge. Using the Monte Carlo technique, our model can accurately simulate combined heat transfer in complex city geometry, accounting for precise thermal and solar radiative heat transfers. Unlike previous models, ours can handle intricate city designs with ease, enabling us to understand the combined effects of conduction, convection and radiation. We have demonstrated the high accuracy of our model, even in detailed 3D city environments. For example, we used it to simulate a heatwave, demonstrating that planting trees can significantly cool streets during the day, although they might slightly reduce night-time radiative cooling. This new tool will empower researchers to improve the coupling of building energy models with urban climate models, and ultimately create cities that are more comfortable and energy-efficient for all.</div
A Menagerie of Wormholes and Cosmologies in the Gravitational Path Integral
No full textInternational audienceWe analyse a variety of Euclidean saddles in the gravitational path integral, with asymptotic AdS boundary conditions, in a class of Einstein-Scalar-Maxwell models. These include single boundary solutions, usual and wineglass wormholes, as well as more exotic (quasi)-oscillatory saddles. Our construction shows how an unbound number of oscillations gets tamed, when flat directions of the potential get lifted. We find several interesting phase transitions between these solutions. The Euclidean wormhole backgrounds can be analytically continued to Lorentzian FLRW universes. Some of them contain an early period of inflation. We delineate the conditions under which they can be the dominant saddles in the gravitational path integral and use them to estimate ratios of probabilities for different cosmological outcomes
On the Impact of the Utility in Semivalue-based Data Valuation
No full text44 pages, 19 figures.International audienceSemivalue–based data valuation uses cooperative‐game theory intuitions to assign each data point a value reflecting its contribution to a downstream task. Still, those values depend on the practitioner’s choice of utility, raising the question: How robust is semivalue-based data valuation to changes in the utility? This issue is critical when the utility is set as a trade‐off between several criteria and when practitioners must select among multiple equally valid utilities. We address this by introducing the notion of a dataset’s spatial signature: given a semivalue, we embed each data point into a lower-dimensional space in which any utility becomes a linear functional, making the data valuation framework amenable to a simpler geometric picture. Building on this, we propose a practical methodology centered on an explicit robustness metric that informs practitioners whether and by how much their data valuation results will shift as the utility changes. We validate this approach across diverse datasets and semivalues, demonstrating strong agreement with rank‐correlation analyses and offering analytical insight into how choosing a semivalue can amplify or diminish robustness
Code de Conduite du GDR-IFM
No full textThis code of conduct was written by the PÉDI committee. On 21 November 2025, the GDR executive committee unanimously approved (with one abstention) a motion that this code has to be accepted by any person participating in any event organised or funded by the GDR.Ce code de bonne conduite a été rédigé par le comité PÉDI. Le comité exécutif du GDR du 21 novembre 2025 a voté à l’unanimité moins un vote blanc le principe selon lequel ce code devait être approuvé par toute personne participant à un événement organisé ou financé par le GDR