Archive ouverte de l'ENAC
Not a member yet
3458 research outputs found
Sort by
Visual Explanations of Differentiable Greedy Model Predictions on the Influence Maximization Problem
No full textInternational audienceSocial networks have become important objects of study in recent years. Social media marketing has, for example, greatly benefited from the vast literature developed in the past two decades. The study of social networks has taken advantage of recent advances in machine learning to process these immense amounts of data. Automatic emotional labeling of content on social media has, for example, been made possible by the recent progress in natural language processing. In this work, we are interested in the influence maximization problem, which consists of finding the most influential nodes in the social network. The problem is classically carried out using classical performance metrics such as accuracy or recall, which is not the end goal of the influence maximization problem. Our work presents an end-to-end learning model, SGREEDYNN, for the selection of the most influential nodes in a social network, given a history of information diffusion. In addition, this work proposes data visualization techniques to interpret the augmenting performances of our method compared to classical training. The results of this method are confirmed by visualizing the final influence of the selected nodes on network instances with edge bundling techniques. Edge bundling is a visual aggregation technique that makes patterns emerge. It has been shown to be an interesting asset for decision-making. By using edge bundling, we observe that our method chooses more diverse and high-degree nodes compared to the classical training
EPLO: Free-space optics emulator for satellite ground links
No full textInternational audienceThis paper presents the first results of the EPLOproject (Free-space optical Propagation test bench mimic),dedicated to the modelling and emulation of the atmosphericeffects on a laser link used for ground-satellite communication.In order to test different modulation waveforms and theirrobustness to tropospheric turbulence, an experimental benchhas been developed to reproduce the front phase deformationand the effect of beam spreading. This emulator generatesvarious scenarios of turbulence based on holographic methods.It includes a beam wander effect, as well as amplitudescintillation modelling and phase deformation. The turbulencepatterns driving the bench are calculated using the Rytovapproximation with extended Kolmogorov spectrum. Based onthis approximation, time series are generated from scintillationindices, beam wandering, beam spreading, phase variation, andthe dynamics of the temporal evolution. Those effects areemulated by a DMD (digital micro-mirror device) using Leeholographic methods. A TILBA-ATMO (Cailabs) isimplemented in the reception chain. Based on Cailabs’ Multi-Plane Light Conversion (MPLC) technology, TILBA-ATMOallows the signal to be coupled into a single-mode fiber whilstcorrecting the effects of atmospheric turbulence. This paperpresents the first efficiency results of this experimental setup onamplitude and phase deformations: this allows us to simulateand study wave front deformation when the laser propagatesthrough the troposphere. By coupling the bench with a singlemode fiber, EPLO will allow a complete emulation of an end-to-end optical link
Changes in plant diversity and vegetation structure based on REVEALS reconstructed vegetation in Europe during the Holocene
No full textInternational audienc
Efficient LDPC-Coded CCSK Links for Robust High Data Rates GNSS
No full textInternational audienceGlobal navigation satellite system links may require increased data rates to accommodate future features and needs (e.g., precise positioning, authentication, reduction of time-to-first-fix data). A particular form of M-ary orthogonal modulation designed for direct-sequence spread-spectrum (DSSS) systems, the cyclic code-shift keying (CCSK) modulation, has been proposed for this purpose. This modulation inherently allows noncoherent processing at receiver side and has the potential to improve the energy efficiency of the data link with respect to classical DSSS/BPSK signals. In this article, q-ary (q∈{2,M}) low-density parity-check (LDPC)-based channel coding for M-ary CCSK is analyzed, both in terms of robustness and computational complexity. (q=M)-ary LDPC-coded CCSK is compared to a bit-interleaved binary LDPC-coded CCSK strategy. Though both solutions provide very reliable links for practical decoding algorithms, it is shown that adequately designed bit-interleaved binary LDPC-coded CCSK signals can offer the additional flexibility inherent to bit-interleaved coded modulation (BICM) while remaining competitive from the point of view of both error rate performance and computational complexity. The latter can be adjusted through the use of incomplete iterative demapping schedules. The optimization of this performance/complexity tradeoff is discussed
Using a Quali-Quantitative Modelling Tool to Explore Scenarios for More-Than-Sustainable Design
No full textInternational audienceThe systemic design approach is particularly relevant to imagining and designing more-than-sustainable futures while respecting planetary boundaries. It involves navigating between spatial, organisational, and temporal scales in order to give weight to long-term effects while addressing the urgency of the situation. It is also a question of perceiving interdependencies and underlying structures, as well as anticipating possible rebound effects.Representation in the form of models, such as causal loop diagrams, sometimes makes it easier to understand and communicate a complex situation, as well as to identify leverage areas for action. Causal loop diagrams represent the structure of a system but do not represent the orders of magnitude of its current state, its dynamics, or projections of possible futures.Yet many of the problems we face are physical and quantifiable (e.g. carbon emissions, land artificialisation, etc.), and we need tools to address them. Systems dynamics, a legacy of hard systems thinking, can help us to imagine, compare and combine design leverages over time through simulation. However, stock and flow modelling tools are hardly used by systemic designers, as they require coding and mastery of the principles of system dynamics.In this workshop, we aim to explore the potential of modelling to support more-than-sustainable design. Prior to the workshop, participants are guided through an example application to familiarise themselves with the formalism of quali-quantitative modelling. Through this example, they discover how the interactive modelling tool can be used to explore scenarios and support design and planning decisions.During the workshop, participants practise quali-quantitative modelling on a case study drawn from their own experience or on the case study presented by the organisers. They are invited to discuss their own experiences and the projects for which modelling would have been of interest before discussing the benefits and limitations of modelling in these different cases. The aim is to share the potential of this approach while gathering feedback from the systemic design community. The workshop is, therefore, primarily aimed at systemic designers who have already worked on projects involving quantitative data and are interested in a modelling tool, but it is open to all kinds of participants, including novice ones
Geodesic regression on SE(3) and application to estimation of positions
No full textInternational audienc
De la MDO à la fabrication : cas d'application pour les drones
No full textInternational audienceThis paper explores the concept of coupling an MDO problem directly to a manufacturing process. The design parameters required for various disciplinary analyses are used to automatically generate geometries that comply with the selected manufacturing processes. Such geometries are then manufactured to validate our proposal. This strategy reduces the manufacturing time of unmanned aerial vehicles (UAVs) as the CAD modeling becomes automatic and parametric, also allowing for easier comparison of different geometries. It can also potentially improve the MDO process by closing the information loop with manufacturing constraints. We employ the Engineering Sketch Pad (ESP) and leverage from additive manufacturing techniques usually applied in UAVs. The choice of ESP facilitates reproducibility, as it is an operating system agnostic open source CAD, and gradient based optimization, due to its capability of computing gradients of the geometric outputs with respect to the design parameters. The manufacturing processes of wings and propellers are addressed. For the same representative wing geometry, we present different modeling strategies for mass and inertia prediction and for manufacturing using 3D printing. ESP is also employed to predict wing mass and inertia. The maximum difference in weight between the manufactured wing and its predicted value using ESP is roughly 7%. The propeller is 3D printed in resin with stereolithography (SLA) technique, and is defined by means of its airfoils, radius, and chord and twist angle distributions. We conclude about the applicability of the presented strategy, as well as its potential and limitations. All the scripts are shared with the community so researchers can apply it to validate their own MDO problem
Pullback bundles and the geometry of learning
No full textInternational audienceExplainable Artificial Intelligence (XAI) and acceptable artificial intelligence are active topics of research in machine learning. For critical applications, being able to prove, or at least to ensure with a high probability the correctness of algorithms is of utmost importance. In practice, however, few theoretical tools are known that can be used for this purpose. Using the Fisher Information Metric (FIM) on the ouput space yields interesting indicators in both the input and parameter spaces, but the underlying geometry is not yet fully understood. In this work, a approach based on the pullback bundle, a well-known trick for describing bundle morphisms, is introduced and applied to the encoder-decoder block. With constant rank hypothesis on the derivative of the network with respect to its inputs, a description of its behaviour is obtained. Further generalization is gained through the introduction of the pullback generalized bundle that takes into account the sensitivity with respect to weights
Equation-Directed Axiomatization of Lustre Semantics to Enable Optimized Code Validation
No full textInternational audienceModel-based design tools like SCADE Suite and Simulink are often used to design safety-critical embedded software. Consequently, generating correct code from such models is crucial. We tackle this challenge on Lustre, a dataflow synchronous language that embodies the concepts that base such tools. Instead of proving correct a whole code generator, we turn an existing compiler into a certifying compiler from Lustre to C, following a translation validation approach.We propose a solution that generates both C code and an attached specification expressing a correctness result for the generated and optionally optimized code. The specification yields proof obligations that are discharged by external solvers through the Frama-C platform