HAL-Polytechnique
Not a member yet
51406 research outputs found
Sort by
Hot off the Press: Proven Runtime Guarantees for How the MOEA/D Computes the Pareto Front From the Subproblem Solutions
No full textInternational audienceThe decomposition-based multi-objective evolutionary algorithm (MOEA/D) does not directly optimize a given multi-objective function f but instead optimizes N + 1 single-objective subproblems of f. We analyze for the first time how the MOEA/D with only standard mutation operators computes the whole Pareto front of the OneMinMax benchmark when the g-optima are a strict subset of the Pareto front. For standard bit mutation, we prove an expected runtime of O(nN log n + nn/(2N)N log n) function evaluations. For the second phase, when the algorithm start with all g-optima, we prove an Ω(n1/2(n/N+1)√N2-n/N expected runtime.For power-law mutation with exponent β ∈ (1, 2), we prove an expected runtime of O(nN log n + nβ log n) function evaluations. The O(nβ log n) term from the second phase is independent of the number of subproblems N, leading to a huge speedup over standard bit mutation. In general, our bound for power-law suggests that the MOEA/D performs best for N = O(nβ-1), resulting in an O(nβ log n) bound
A Selective and Sensitive Method for Colistin Detection by G-Quadruplex Ligand Competition
No full textInternational audienceColistin (COL) is a widely used antibiotic and is quite often used as a last-resort treatment option for treating multidrug-resistant Gram-negative bacterial infections. Due to its widespread use, COL accumulates in nature, which represents a novel ecological and health threat. However, there is currently no rapid and specific method available for titrating COL levels in collected samples. Herein, we report a simple chemiluminescence detection method based on the specific interaction between COL and a parallel G-quadruplex (G4). To this end, we exploit the catalytic properties of the G4/hemin DNAzyme, which is able to oxidize substrates to provide a readily monitored readout. The stronger affinity of G4 for COL versus hemin allows for the inactivation of the G4/hemin DNAzyme, which is used herein to quantify COL in solution. Through a series of optimizations, we identified the best G4 sequence (F3TC), oxidation substrate (luminol), and experimental conditions, which allow for the detection of COL over a broad concentration window, from 0.5 to 2,500 ng/mL, with a detection limit of 0.4 ng/mL and excellent selectivity against other antibiotics. Compared with existing methods, the proposed approach provides a simpler and label-free quantification of COL, which might serve as a valuable standard method for antibiotic detection, whose use was validated under real conditions herei
Le rôle des plateformes numériques sur le marché du travail : enjeux d'adoption, impact sur l'emploi et équité algorithmique
No full textThis thesis seeks to deepen the understanding of digital job search tools by shedding light on the opportunities they create and on the challenges they may pose. It focuses on three main aspects: the adoption of job search tools, their potential benefits in terms of employment outcomes, and the risks tied to their usage. All chapters are leveraging detailed data provided by the French Public Employment Service, France Travail.The first chapter explores the determinants of job platform adoption by testing various email communication strategies designed to encourage their usage. These interventions, informed by behavioral economics and psychological literature and developed in collaboration with caseworkers, include informational components, varying levels of assistance, or a combination of these levers. In terms of methodology, the study employs adaptive randomized controlled trials, an innovative public policy experimentation methods. These methods are designed to identify the most effective levers to facilitate user engagement during the experiment. Additionally, the chapter explores how different treatments may affect jobseekers based on their individual characteristics, with the goal to construct a personalized communication strategy. The results show that no intervention significantly outperformed the minimal control email on average, and that a personalized strategy showed only marginal improvements in increasing platform adoption.The second chapter extends the first by investigating the medium-term impact of digital platforms. It introduces a theoretical framework to understand the mechanisms through which these digital tools can improve individuals' prospects in the labor market. Based on empirical data collected in the first chapter, it examines how the increased usage of the job platform, alongside traditional job search channels, affects the behavior of job seekers. The findings align with the predicted theoretical results showing that jobseekers who engaged more with the digital tools received more job propositions from recruiters, reduced their job search efforts, and experienced a slight increase in stable employment.The third chapter focuses on the impact of recommendation algorithms, which have become a key tool on digital job platforms for personalizing job suggestions based on user profiles and preferences. It conducts a detailed audit of a recommender system, trained on hires, to identify potential treatment disparities among groups of jobseekers, focusing on gender gaps. Drawing from labor economic literature, the chapter proposes a measure of fairness in recommendation and highlights the importance of considering gaps in terms of job ad characteristics, possibly controlling for jobseekers' qualifications and preferences, when assessing fairness in recommendations. The findings indicate that the studied algorithm suggests jobs with differing characteristics to men and women, even when controlling for jobseekers qualifications and preferences. However, these gaps are not greater than those observed in real hires. To address gaps in practice, the chapter also proposes an efficient post-processing correction method that can be easily implemented on a large scale. It evaluates the correction's impact on the trade-off between performance and fairness and demonstrates that the correction allows to reach the fariness objectif as per the proposed measure with minimal impact on the algorithm performance.Cette thèse vise a mieux comprendre l'impact des outils numériques de recherche d'emploi sur le marché du travail en mettant en lumière les nouvelles opportunités qu'ils offrent aux demandeurs d'emploi ainsi que les défis qu'ils posent. Elle se concentre sur trois aspects : l'adoption de ces outils de recherche d'emploi, leurs gains potentiels en termes de retour à l'emploi, et les risques liés à leur utilisation. Tous les chapitres s'appuient sur des données détaillées fournies par le Service Public de l'Emploi Français, France Travail.Le premier chapitre explore les déterminants de l'adoption des plateformes d'emploi en testant diverses stratégies de communication par e-mail conçues pour encourager leur utilisation. Ces interventions, basées sur la littérature en psychologie et en économie comportementale et développées en collaboration avec des professionnels du terrain, comprennent des composants informatifs, différents niveaux d'aide à la réalisation, ou une combinaison de ces leviers. La méthodologie repose sur des expériences randomisées contrôlées adaptatives, conçue pour identifier les leviers les plus efficaces pour accroître l'engagement des utilisateurs pendant l'expérience. Le chapitre examine également comment ces stratégies affectent les demandeurs d'emploi en fonction de leurs caractéristiques individuelles, afin de développer une communication personnalisée. Les résultats montrent qu'aucune intervention n'a significativement donné de meilleurs résultats en moyenne que l'intervention dite de contrôle. De plus, une stratégie de communication personnalisée n'a montré que des améliorations marginales dans l'adoption de la plateforme d'emploi comparée à une stratégie aléatoire.Le deuxième chapitre prolonge cette analyse en examinant l'impact à moyen terme des plateformes numériques d'emploi. Il introduit un cadre théorique pour comprendre les mécanismes par lesquels ces outils peuvent améliorer les perspectives professionnelles des individus. Il analyse, sur la base des données du premier chapitre, comment l'utilisation de la plateforme d'emploi, en parallèle de canaux traditionnels de recherche, influence le comportement des demandeurs d'emploi et des recruteurs. Les résultats confirment les prévisions théoriques du modèle : les demandeurs d'emploi utilisant les outils numériques reçoivent davantage de propositions d'emploi de la part des recruteurs, réduisent leurs efforts de recherche, et augmentent légèrement leurs chances de retrouver un emploi stable.Le troisième chapitre se concentre sur l'impact des algorithmes de recommandation, devenus des outils clés pour personnaliser les suggestions d'emploi sur plateformes d'emploi numériques. Il réalise un audit détaillé d'un système de recommandation, entraîné sur les embauches, afin d'identifier les disparités de traitement entre groupes de demandeurs d'emploi, en mettant l'accent sur les écarts de genre. En s'appuyant sur la littérature en économie du travail, le chapitre propose une mesure de l'équité des recommandations et souligne l'importance de la prise en comptes des écarts en termes de caractéristiques des offres d'emploi recommandées, en contrôlant potentiellement des qualifications et des préférences des demandeurs d'emploi. Les résultats montrent que l'algorithme étudié recommande des emplois différents aux hommes et aux femmes, même en contrôlant des qualifications et préférences objectives des demandeurs d'emploi. Cependant, ces écarts ne sont pas plus importants que ceux observés dans les recrutements effectifs. Pour remédier à ces écarts en pratique, le chapitre propose une méthode de correction post-traitement efficace et facilement déployable à grande échelle. Il évalue l'impact de cette correction sur l'arbitrage entre performance et équité des recommandations et démontre que la correction permet d'atteindre les objectifs d'équité selon la mesure proposée avec un impact minimal sur la performance de l'algorithme étudié
Search for the decay
No full textInternational audienceA search for the decay is made using collision data collected with the LHCb detector at centre-of-mass energies of 7, 8 and 13 TeV, corresponding to an integrated luminosity of fb. No significant signal is observed, and an upper limit on the branching fraction of at confidence level is set. This result supersedes the previous LHCb study and improves the upper limit by a factor of two
Search for resonances decaying to photon pairs with masses between 4.9 and 19.4 GeV
No full textInternational audienceA search is presented for axion-like particles (ALPs) with masses between 4.9 and 19.4 GeV decaying to a pair of photons, using proton-proton collisions collected with the LHCb detector during 2018 at a centre-of-mass energy of 13 TeV, corresponding to an integrated luminosity of 2.1 fb. The same strategy and sample is used to search for the decays of the , and mesons into photon pairs. No significant excess is found. Upper limits on the photon-pair branching fraction times the cross-section of ALP production are determined as a function of the ALP mass. Limits on the branching fractions of the beauty states are determined to be , , and at 95 % confidence level
Precision measurement of the baryon lifetime
No full textInternational audienceA sample of collision data, corresponding to an integrated luminosity of 5.4 fb and collected by the LHCb experiment during LHC Run 2, is used to measure the ratio of the lifetime of the baryon to that of the baryon, . The value is obtained, where the first uncertainty is statistical and the second systematic. This value is averaged with the corresponding value from Run 1 to obtain . Multiplying by the known value of the lifetime yields , where the last uncertainty is due to the limited knowledge of the lifetime. This measurement improves the precision of the current world average of the lifetime by about a factor of two, and is in good agreement with the most recent theoretical predictions
Federated Majorize-Minimization: Beyond Parameter Aggregation
No full textThis paper proposes a unified approach for designing stochastic optimization algorithms that robustly scale to the federated learning setting. Our work studies a class of Majorize-Minimization (MM) problems, which possesses a linearly parameterized family of majorizing surrogate functions. This framework encompasses (proximal) gradient-based algorithms for (regularized) smooth objectives, the Expectation Maximization algorithm, and many problems seen as variational surrogate MM. We show that our framework motivates a unifying algorithm called Stochastic Approximation Stochastic Surrogate MM (SA-SSMM), which includes previous stochastic MM procedures as special instances. We then extend SA-SSMM to the federated setting, while taking into consideration common bottlenecks such as data heterogeneity, partial participation, and communication constraints; this yields FedMM. The originality of FedMM is to learn locally and then aggregate information characterizing the surrogate majorizing function, contrary to classical algorithms which learn and aggregate the original parameter. Finally, to showcase the flexibility of this methodology beyond our theoretical setting, we use it to design an algorithm for computing optimal transport maps in the federated setting
Cross-species immune activation and immunobiotics: a new frontier in vector-borne pathogen control
No full textInternational audienceThe persistent global burden of vector-borne diseases (VBDs) needs innovative control strategies, as traditional methods are compromised by acaricides and drug resistance and variable vaccine efficacy. We propose a dual-action strategy using cross-species immune activation: human microbiota triggers the production of natural antibodies that directly target pathogens in the host and modulate vector immunity by interacting with vector microbiota. The human microbiota also modulates cytokine responses, enhancing immune defenses in both host and vector. These mechanisms can be further optimized by identifying immunobiotics – specific gut microbes that stimulate protective immune responses against VBDs. This approach offers a sustainable framework to bridge the gap between host and vector immunity, introducing a novel method to combat VBDs
Perfectly Matched Layers implementation for E-H fields and Complex Wave Envelope propagation in the Smilei PIC code
No full textInternational audienceThe design of absorbing boundary conditions (ABC) in a numerical simulation is a challenging task. In the best cases, spurious reflections remain for some angles of incidence or at certain wavelengths. In the worst, ABC are not even possible for the set of equations and/or numerical schemes used in the simulation and reflections can not be avoided at all. Perflectly Matched Layers (PML) are layers of absorbing medium which can be added at the simulation edges in order to significantly damp both outgoing and reflected waves, thus effectively playing the role of an ABC. They are able to absorb waves and prevent reflections for all angles and frequencies at a modest computational cost. They increase the simulation accuracy and negate the need of oversizing the simulation usually imposed by ABC which normally leads to a waste of computational resources and power. In this paper, a uniform derivation of PML for finite-difference time-domain (FDTD) schemes and various geometries in Particle-In-Cell (PIC) codes is presented for Maxwell's equations and, for the first time, extended to the full envelope wave equation. An implementation of these methods in the open source PIC code Smilei is proposed and benchmarked
Heavy-Heavy-Light Asymptotics from Thermal Correlators
No full textInternational audienceWe revisit the calculation of spectral densities and heavy-heavy-light (HHL) operator product expansion (OPE) coefficients in three-dimensional conformal field theories using thermal one-point functions on . A central element of our analysis is a new inversion formula for one-point functions which is derived via Casimir differential equations. We develop systematic expansions of the spectral density and HHL OPE coefficients in the regime of large . We validate our analytic tools by comparing the results with the partial wave expansions of thermal one-point functions in free field theories. The algorithms developed for these expansions make full use of Casimir recursion relations, thereby extending their applicability into the heavy exchange regime. In the end, we observe excellent agreement with our analytic predictions and an improvement of up to three orders of magnitude compared to all previous leading order estimates of the CFT data even for moderate values of