Archive ouverte HAL-LAAS
Not a member yet
12189 research outputs found
Sort by
Algorithmic curation of News on YouTube: Evidence from the 2022 French Presidential Campaign
No full textDebate is growing over how algorithmic recommendations influence news visibility, particularly regarding interest and ideological bias. This study examines YouTube's News Recommender System (NRS), focusing on the French "News" section of the homepage through a large-scale audit using automated browsing agents.Our findings show that the NRS prioritises platform-native creators over established news outlets, favouring content that aligns with YouTube's features rather than traditional editorial standards. Politically charged, opinion-based videos, especially those from prominent figures affiliated with extreme political parties, receive ongoing algorithmic promotion. While centrist and moderate political figures remain underrepresented, the algorithm boosts their visibility once users interact with this type of content. This two-part mechanism, which amplifies already prominent content and appears to overcompensate for rare content, does not just reflect engagement-based optimisation, but is driven by the algorithm's tendency to maintain coherence in a highly unbalanced content landscape. However, this compensatory logic does not counteract the algorithm's broader tendency to promote content from radical political figures while marginalising institutional news outlets. Through this process, the NRS actively reshapes news exposure within the "News" section, privileging political expression over journalistic authority and reproducing structural hierarchies of visibility
Long-Horizon Task and Motion Planning with Learning-Based Geometric Reasoning
No full textWe propose TAMP-LGR, a novel task and motion planning framework that integrates learned geometric reasoning into a search-based planning process. The approach addresses one of the main challenges in Task and Motion Planning (TAMP), which is the high computational cost of repeated geometric feasibility checks. To address this, we introduce Geometric Reasoning Networks (GRN), a graph neural network that predicts action and grasp feasibility while identifying causes of infeasibility such as inverse kinematic limitations or grasp obstructions. These predictions are used to build a set of geometric constraints that guide the symbolic planner towards feasible solutions. TAMP-LGR combines two complementary heuristics: a feasibility heuristic that prioritizes geometrically valid actions, and an informed cost-to-goal heuristic that leverages infeasibility constraints to reduce the combinatorial complexity of planning. Extensive experiments on challenging single-and multi-robot manipulation benchmarks demonstrate that TAMP-LGR substantially outperforms existing TAMP methods in both efficiency and success rate, and solves large-scale problems that remain challenging for state-of-the-art approaches. We further show how the framework can be extended to mobile manipulators, maintaining its scalability and generalization capabilities
Identificación de Parámetros de Baterías de Ión-Litio: Un Estudio Comparativo de Diversos Modelos y Técnicas de Optimización para el Modelado de Baterías
No full textInternational audienceThis work presents a comparative study of optimization techniques for parameter identification in equivalent electrical models of lithium-ion batteries. The 2RC model is applied to a set of twelve batteries using four publicly available datasets obtained from well-established research institutions. The methodology is structured in four stages: first, the 2RC model is selected due to its balance between physical interpretability and computational simplicity; second, experimental charge-discharge cycle data are collected; third, various optimization techniques are applied with the aim of minimizing the error between the experimental data and the response estimated by the model; and finally, accuracy is evaluated using the mean squared error, while computational efficiency is assessed through execution time. Traditional, metaheuristic, and bio-inspired optimization methods are considered, including least squares optimization, particle swarm optimization, simulated annealing, and several nature-inspired variants. It is demonstrated that bio-inspired techniques achieve greater accuracy than traditional methods, without a significant increase in computational cost. In particular, particle swarm optimization shows superior performance in terms of precision and robustness against local minima. It is concluded that the integration of advanced optimization strategies significantly enhances the fidelity of equivalent electrical models, which is essential for accurate estimation of internal states such as state of charge, aging, and service life in lithium-ion batteries used in electric vehicles and aerospace systems.En este trabajo se lleva a cabo un estudio comparativo de técnicas de optimización para la identificación de parámetros en modelos eléctricos equivalentes de baterías de ion-litio. Se emplea el modelo 2RC sobre un conjunto de doce baterías, utilizando cuatro bases de datos públicas provenientes de centros de investigación reconocidos. La metodología se estructura en cuatro etapas: en primer lugar, se selecciona el modelo 2RC por su equilibrio entre precisión y simplicidad computacional; en segundo lugar, se recopilan datos experimentales de ciclos de carga y descarga; en tercer lugar, se aplican diversos métodos de optimización con el objetivo de minimizar el error entre los datos experimentales y los resultados estimados por el modelo; y finalmente, se evalúan la precisión, mediante el error cuadrático medio, y la eficiencia computacional, mediante el tiempo de ejecución. Se consideran métodos tradicionales, metaheurísticos y bioinspirados, entre ellos la optimización por mínimos cuadrados, el algoritmo de enjambre de partículas, el recocido simulado y diversas variantes inspiradas en procesos naturales. Se evidencia que las técnicas bioinspiradas permiten alcanzar una mayor precisión que los métodos tradicionales, sin un aumento significativo en el costo computacional. En particular, la optimización por enjambre de partículas muestra un desempeño superior en cuanto a exactitud y robustez frente a mínimos locales. Se concluye que la incorporación de estrategias de optimización avanzadas mejora significativamente la fidelidad de los modelos eléctricos equivalentes, lo que resulta fundamental para una estimación más precisa del estado de carga, el envejecimiento y la vida útil de baterías en aplicaciones críticas, tales como vehículos eléctricos y sistemas aeroespaciales
Cycle Patterns and Mean Payoff Games
No full textWe introduce the concept of a cycle pattern for directed graphs as functions from the set of cycles to the set {-,0,+}. The key example for such a pattern is derived from a weight function, giving rise to the sign of the total weight of the edges for each cycle. Hence, cycle patterns describe a fundamental structure of a weighted digraph, and they arise naturally in games on graphs, in particular parity games, mean payoff games, and energy games. Our contribution is threefold: we analyze the structure and derive hardness results for the realization of cycle patterns by weight functions. Then we use them to show hardness of solving games given the limited information of a cycle pattern. Finally, we identify a novel geometric hardness measure for solving mean payoff games (MPG) using the framework of linear decision trees, and use cycle patterns to derive lower bounds with respect to this measure, for large classes of algorithms for MPGs
Exploring Plasma Cell-Free DNA Structure for Agnostic Biomarker Identification
No full textInternational audienceAbstractBACKGROUND : Cell-free DNA (cfDNA) is a promising biomarker, offering both disease-specific insights (e.g., mutations) and systemic, agnostic information (e.g., fragmentomics). Our team has recently pioneered the BIABooster technology, which, to the best of our knowledge, represents the most sensitive DNA sizing technology available (1 fg/µL). This innovation enables the characterization of cfDNA in plasma without purification within minutes. Utilizing this technology, we have demonstrated that the cfDNA size profile serves as an agnostic biomarker, differentiating advanced lung cancer patients from healthy controls.AIM : We hypothesize that the analysis of cfDNA size profiles can be further enhanced through a deeper understanding of cfDNA structure in plasma. CfDNA is known to associate with various biological solutes in the bloodstream, including lipoproteins and extracellular vesicles (EVs). These interactions are likely modulated in disease conditions but remain poorly characterized in terms of their molecular diversity and functional relationship to cfDNA stability.RESULTS : In this study, we present an analytical workflow for plasma fractionation based on size, density, or combined size-density parameters. We detail the cfDNA composition of the resulting fractions, revealing that in healthy individuals, cfDNA is not associated with EVs. Instead, it is found within small, non-EV particles whose density correlates with cfDNA size. Applying this analytical workflow to cancer patient samples, we suggest that the structure of cfDNA is altered in cancer, providing new avenues for biomarker discovery and clinical application
Exploiting Term Sparsity in Symmetry-Adapted Basis for Polynomial Optimization
No full textInternational audiencePolynomial optimization problems are infinite-dimensional, nonconvex, NP-hard, and are often handled in practice with the moment-sums of squares hierarchy of semidefinite programming bounds. We consider problems where the objective function and constraint polynomials are invariant under the action of a finite group. The present paper simultaneously exploits group symmetry and term sparsity in order to reduce the computational cost of the hierarchy. We first exploit symmetry by writing the semidefinite matrices in a symmetry-adapted basis according to an isotypic decomposition. The matrices in such a basis are block diagonal. Secondly, we exploit term sparsity on each block to further reduce the optimization matrix variables. This is a non-trivial extension of the term sparsity-based hierarchy related to sign symmetry that was introduced by two of the authors. Our method is compared with existing techniques via benchmarks on quartics with dihedral, cyclic and symmetric group symmetry
Optimizing dimensional accuracy in two-photon polymerization: Influence of energy dose and proximity effects on sub-micrometric fiber structures
No full textInternational audienc
Anonymization Did Not Fail: Misconceptions and Overstatements on Data Anonymization Failures
No full textInternational audienceSeveral authors have claimed the “failure of anonymization,” despite over 50 years of research. We review privacy leaks reported over the past decades and conclude they were due to nonexistent or inadequate anonymization, rather than a lack of robust anonymization methods
Self-mixing interferometry system for in-vitro flow mapping of retinal arteriolar network
No full textInternational audienceSelf-Mixing Interferometry (SMI) is an optical sensing technique that enables the creation of compact, all-in-one optical sensors for high-resolution measurements, making it an attractive tool for flowmetry applications, such as velocity mapping in microfluidic systems. Most research in this area has focused on artificial rectangular or circular channels, which do not fully replicate in vivo-like structures. This study demonstrates the application of SMI for velocity mapping in microchannels designed to mimic the retinal arteriolar network. These microchannels were fabricated using backside lithography, a novel technique that produces semi-rounded geometries closely resembling in vivo conditions. A high-resolution SMI system was developed, achieving accurate velocity measurements with a spatial resolution of 1 μm for detailed flow profiles, as well as faster scans at lower resolutions for global flow patterns. The system's ability to reconstruct velocity maps and track flow variations within an artificial vascular network highlights the potential of SMI sensors for use in more complex, in vivo-like applications
A floating inverted pendulum model for analysing the pitch stability of offshore wind turbines
No full textInternational audienceThis work presents a reduced-order dynamic model for floating offshore wind turbines (FOWTs), formulated as a floating inverted pendulum with explicit representation of the platform's pitch motion. The model incorporates hydrostatics, added-mass and radiation effects from potential-flow hydrodynamics, and an equivalent mooring system with both stiffness and viscous contributions. A nonlinear equation of motion is derived and complemented by state-space and linearised representations, enabling both dynamic assessment and control-oriented applications. Environmental excitations are introduced from stochastic wind-wave realisations generated in QBlade, ensuring direct comparison with high-fidelity six-degree-of-freedom simulations. Validation is performed under three representative offshore conditions, ranging from mild to severe seas, showing close agreement in pitch response statistics and spectral content. Results indicate that while the reduced-order model slightly underestimates variability at harsher sea states, it reliably reproduces dominant frequencies and overall response trends. Beyond validation, the framework is extended to include active mooring line control based on a Linear Quadratic Regulator (LQR), and further explored through a control co-design (CCD) methodology in which structural parameters and controller gains are optimised simultaneously. Overall, the proposed model provides a physically consistent yet computationally efficient tool for early-stage design, stability analysis, and the development of advanced control strategies for floating offshore wind turbines