HAL-Ecole des Ponts ParisTech
Not a member yet
42743 research outputs found
Sort by
Perceptions on the adoption of Free/Open Source Software policies by a Scientific Institution: The case study of the NIH
No full textInternational audienceBackground. As the Open Science context evolves broadly and Scientific Institutions implement Open Access and Open Science policies, it is of interest to observe which are the issues that come into play when these Institutions attempt to contribute to make their research outputs visible, accessible and reusable. The present work reflects on the problematic behind the establishment and adoption of policies regarding the dissemination of Research Software (RS) as Free/Open Source Software (FOSS) in a Scientific Institution.Method. In this work we begin by establishing an initial, general framework regarding RS and FOSS fundamental concepts and terminology. Then we include the initial list of questions for the NIH-RFI, followed by our detailed answers, argued in the context of the provided framework.Results. Besides presenting this large collection of answers, we also include diverse reflections on what we have observed and learned in the context of the analysis of the NIH-RFI, arguing that it also provides us with the opportunity to explore how the previously published work on RS and Open Science can be envisioned in the lens of the definition and adoption of RS sharing and disseminating policies by any Scientific Institution or research community.Conclusions. We have detailed here a foundational framework, showing how it could be considered as the key tool for providing a systematic reply to the RS and FOSS issues arising in the NIH-RFI. This framework can be easily adapted to more general contexts, and, thus, could be of potential interest for other scientific organizations, as well as for scientific communities with an important RS production, like the Computer Algebra and the Symbolic Computation ones, aiming to clarify and upgrade their policies and practices regarding this kind of production.</div
The late-stage training dynamics of (stochastic) subgradient descent on homogeneous neural networks
No full textInternational audienceWe analyze the implicit bias of constant step stochastic subgradient descent (SGD). We consider the setting of binary classification with homogeneous neural networks -a large class of deep neural networks with ReLU-type activation functions such as MLPs and CNNs without biases. We interpret the dynamics of normalized SGD iterates as an Euler-like discretization of a conservative field flow that is naturally associated to the normalized classification margin. Owing to this interpretation, we show that normalized SGD iterates converge to the set of critical points of the normalized margin at late-stage training (i.e., assuming that the data is correctly classified with positive normalized margin). Up to our knowledge, this is the first extension of the analysis of the Lyu and Li (2020) on the discrete dynamics of gradient descent to the nonsmooth and stochastic setting. Our main result applies to binary classification with exponential or logistic losses. We additionally discuss extensions to more general settings
Conflict Reporting in the Digital Age
No full textThe digital age has fundamentally reshaped international traditional news reporting of foreign conflicts. Online posts by first-hand witnesses influence the volume, slant, content, informativeness, and tone of international media conflict coverage. Exploiting exogenous variation from local internet outages in the conflict zone, we show that U.S. TV networks air more stories about the Israeli-Palestinian conflict when ordinary users caught in the conflict have internet access. These stories differ in both slant and focus: in the absence of internet disruptions, news coverage is more sympathetic toward the conflict side experiencing higher casualties, places greater emphasis on civilian suffering, and devotes less attention to foreign policy and electoral considerations. This coverage is also more emotional, includes more local detail, references social media platforms more frequently, and is more homogeneous across ideologically diverse networks. Importantly, these conflict stories generate greater viewer engagement
Participatory downscaling of global SSP–RCP scenarios to local fisheries social–ecological systems
No full textInternational audiencePolicymakers and managers need projections of the trajectories of change in human societies and natural ecosystems to plan actions to address the potential consequences of global change. There are many projections at the global scale, but these need to be downscaled to local marine social–ecological systems to plan targeted interventions. We aimed to provide policy-relevant projections of the consequences of climate change and socio-economic orientations for the fisheries of the French North Sea and the Mediterranean. Our approach consisted of a downscaling of global scenarios of IPCC Shared Socioeconomic Pathways (SSPs) and Representative Concentration Pathways (RCPs) for the 2100 horizon. We engaged with fisheries managers, policymakers, scientific experts, and NGOs, and co-constructed four contrasting scenarios through interdisciplinary participatory workshops. Our approach ensured strong anchoring in regional, national, and local contexts and provided references to fisheries scenarios. We then conducted a novel comparative analysis of SSP–RCP scenario narratives from different local marine social–ecological systems in the same country. Across all scenarios, 83% of the themes highlighted by the stakeholders were common between the two social–ecological systems. However, within the themes, at a more detailed level, only 30% of the narrative elements were mentioned in both systems, reflecting the specificity and richness of these downscaled scenarios and their relevance to local management. We also integrated disruptive changes and emotional aspects by conducting a novel standardized emotion analysis using an artificial intelligence language model. Overall, these downscaled scenarios identified pathways to be avoided, as well as potential actions for transformative change, mitigation, and adaptation to global change at the local scale. Our protocol allowed us to address some criticisms of SSP–RCP scenarios by helping to bridge science–policy gaps and to focus more on human–nature relationships
Impact de la politique commerciale américaine sur les économies africaines
No full textLe nouveau régime tarifaire des États-Unis va bouleverser d’innombrables relations commerciales ainsi que le système financier mondial, entraînant de fortes volatilités sur les marchés et un ralentissement de la croissance. Parmi ce lot de mauvaises nouvelles, on peut citer l’impact négatif sur les pays d’Afrique qui bénéficiaient d’un accès en franchise de droits aux marchés américains pour de nombreux produits. Cependant, le risque pour les économies africaines viendra davantage de la volatilité des prix de matières premières qui pourrait mettre en péril l’écosystème commercial africain et leur équilibre extérieur, et des chocs sur les taux souverains africains
How Do Immigrants Promote Exports?
No full textInternational audienceHow do immigrants promote exports? To answer this question we propose an empirical framework allowing to disentangle the role of migration networks that operate at a bilateral level from that of productivity channels (knowledge diffusion and increased workforce diversity) that operate at the aggregate level. We find evidence supporting both, at the extensive as well as at the intensive margin. The results are robust to using various IV strategies. While richer countries’ exports tend to benefit more from immigrants’ diversity (especially in sectors characterized by complex production processes), developing countries benefit from knowledge diffusion more
From macro to micro: Large exporters coping with global crises
No full textInternational audienceUsing monthly firm-level exports and imports over 1993–2020, we uncover four facts: (i) deviations of large exporters from the average growth rate explain a large share of aggregate fluctuations; (ii) an important source for these deviations is the top exporters’ higher loadings on common shocks; (iii) the stronger reaction of the top 1% exporters to the GFC and Covid crises contributed to the export collapses; (iv) a higher elasticity to large demand shocks, not a different exposure to global value chain shocks, contributes to this stronger reaction. The results show that idiosyncratic reactions of large firms to common shocks matter for aggregate export fluctuations, and are especially relevant for the trade collapses of the 2008/2009 crisis and the Pandemic
Adaptive and interoperable federated data spaces: an implementation experience
No full textInternational audienceAs modern smart cities increasingly rely on federated IoT-enabled data spaces to support their operations, the dynamic and heterogeneous nature of these environments introduces significant challenges. Variations in data models, protocols, and policies hinder seamless federation and collaboration. To address these challenges, we present the prototype implementation of CCDUIT, a modular, scalable, and adaptive software overlay architecture for cross-federation collaboration. This architecture dynamically adapts to evolving data exchange requirements, including protocol variations, data model discrepancies, and policy constraints, ensuring continuity and efficiency in data discovery and exchange. The prototype uses context-aware mechanisms to autonomously adjust operations, dynamically reconfigure collaborations, enforce data exchange policies, and optimize data-sharing pathways. We validate the prototype using real-world data spaces operating to address barriers to climate change mitigation and energy transition. Experimental results highlight the ability of the architecture to adapt for diverse federation-specific needs with minimal operational overhead and respond to dynamic policy changes. We release our prototype as open source, enabling the community to explore, extend, and adapt it to their own needs
Le dilemme climat/développement dans les négociations internationalessur le climat : Histoire d’un réductionnisme climato‐centré(1990‐2022)
No full textTraduction de la thèse 'Le dilemme climat/développement dans les négociations internationales sur le climat : Histoire d’un réductionnisme climato‐centré (1990‐2022)' soutenue à l'EHESS par Christophe Cassen le 18 octobre 2023The dissertation examines the persistent climate‐centric approach of the international climate negotiations and its related tension with development issues. It analyzes this tension through the coevolution between the main policy instruments adopted within the UNFCCC and a climate policy engineering, in which techno‐economic modeling occupies a distinctive position. Chapter 1 presents an interdisciplinary approach to analyzing these co‐evolutions and mobilises the concept of path dependency. It draws on a variety of sources, including academic and grey literature, CIRED archives and stakeholder interviews. The second part of the dissertation (chapters 2 and 3) examines two dimensions of a path‐dependent process that has led to a climate‐centric approach, including the coordination of negotiations around targets, from the Kyoto Protocol to the Paris Agreement, and persistent equity concerns about a burden‐sharing approach. The contribution to this path dependent process of energy economic experts from North countries, primarily in international fora and organisations is examined at four key moments: i) the European ecotax project (1990‐1992), ii) the negotiations on the flexibility mechanisms of the Kyoto Protocol (1995‐2001), iii) the adoption of mechanisms dedicated to South countries prior to Bali (NAMAs) (2001‐2007), and iv) theinstitutionalisation of the 2°C objective (2007‐2014). The third part analyzes the role of Integrated Assessment Models (IAMs) as a specific facet of the climate policy engineering in the climate‐centric approach of the negotiations. Chapter 4 shows that the reorientation of the policy agenda towards energy issues in the 80s had a significant impact on modeling efforts, particularly on the new IAMs that emerged following the institutionalization of the climate change issue. Chapter 5 shows how IAMs thenbecame central in producing scenarios consistent with the 2°C target, based on a 'techno‐centric' approach. To that end, the IAM community relied on a repertoire of modeling practices and close collaboration with IPCC Group III (Chapter 6). The fourth part describes a fifth post‐Paris moment first characterized by a persistent gap between the long‐term transition scenarios generated by IAMs, in line with the IPCC AR5, and the short‐term challenges of accelerating the low‐carbon transition, as discussed in Chapter 7. Chapter 8 also highlights the growing role of modeling communities in the construction of national decarbonisation strategies and the organiszational and scientific challenges they face. The dissertation concludes that the incremental dynamics of change in the UNFCCC policy instruments have co‐evolved with opposite conceptions of the climate‐development nexus within a climate policy engineering, due to both strategies of alignment with the negotiating agenda and internal dynamics (techno‐centric reductionism, North‐South asymmetry, community effect). The conclusion calls for a more in‐depth examination of recent change and continuity within this technoeconomicexpertise and of its interfaces within the current climate negotiations.La thèse analyse la tension persistante entre climato‐centrisme et développement au sein des négociations climat, au travers des co‐évolutions entre les modalités de coordination et une Climate Policy Engineering au sein de laquelle la modélisation technico‐économique occupe une position singulière.Le chapitre 1 présente une approche interdisciplinaire qui analyse ces co‐évolutions en se centrant sur la notion de dépendance au sentier et s’appuie sur un panel varié de sources : littérature académique et grise, archives du CIRED, interviews d’acteurs.La deuxième partie, dans les chapitre 2 et 3 examine comment cette tension s’exprime au sein de deux dimensions d’un processus dépendant au sentier : la première à l’égard d’une coordination par les objectifs (quantifiés à Kyoto puis en température à Copenhague et Paris), la seconde vis‐à‐vis des enjeux d’équité, sous l’angle du partage du fardeau des émissions à Kyoto et ensuite autour du financement Nord‐Sud depuis Bali. Cette tension est notamment entretenue par l’empilement institutionnel qui caractérise le changement de paradigme des négociations depuis Bali. Enfin, la contribution de communautés d’experts (économistes‐ingénieurs) de l’énergie en majorité du Nord au sein de fora et d’organisations internationaux est appréhendée dans quatre principaux moments de l’histoire des instruments de coordination : i) le projet d’écotaxe européen (1990‐1992), ii) les négociations autour des mécanismes de flexibilité du Protocole (1995‐2001), iii) l’intégration de mécanismes à destination duSud en amont de Bali (NAMAs) (2001‐2007), iv) l’institutionnalisation de l’objectif 2°C ( 2007‐2014).La troisième partie met notamment en lumière un réductionnisme ‘technico‐centré’ par les modèles intégrés (IAMs) dans la production de scénarios compatibles avec l’objectif 2°C hérité du recentrage de la modélisation technico‐économique dans les années 1980 (chapitre 4). Dans le contexte d’institutionnalisation de l’objectif 2°C, son articulation avec une approche climato‐centrée a relégué au second plan l’évaluation des conditions de bifurcation des trajectoires de développement (chapitre 5).L’ orientation de la modélisation intégrée pendant cette période est étroitement liées aux modalités de structuration de la communauté des IAM (répertoire de pratiques de modélisation et une collaboration étroite avec le groupe III du GIEC) (chapitre 6).Dans une quatrième partie, sont mis en avant les contours d’un 5e moment Post‐Paris marqué d’un côté par un décalage persistant entre le long terme des scénarios de transition proposés par les IAM (dans la lignée du 5e rapport du GIEC) et les enjeux de court terme et d’accélération de l’action (chapitre 7). Le chapitre 8 décentre de son côté la perspective globale de la thèse vers le rôle des assemblages de modélisation dans la construction des stratégies nationales de décarbonation. Le dilemme climat‐développement se joue désormais aussi à ce niveau dans le contexte de mise en oeuvre des objectifs de l’Accord de Paris.La thèse conclut sur l’importance dans la récurrence des tensions climat‐développement des processus d’empilement et d’enchevêtrement institutionnels dans les dynamiques incrémentales de changement à l’oeuvre tout au long de la période. Ceci est allé de pair avec une supersposition de cadrages de l’articulation climat‐développement au sein des moments de co‐évolution avec une Climate Policy Engineering. Elle met enfin en évidence le besoin d’approfondir l’examen des transformations des dispositifsde modélisation à l’échelle nationale et internationale dans le contexte de la mise en oeuvre de l’Accord de Paris et du renforcement de la coopération Nord‐Sud
Applying the Dempster–Shafer Fusion Theory to Combine Independent Land-Use Maps: A Case Study on the Mapping of Oil Palm Plantations in Sumatra, Indonesia
No full textAll data and codes used are available on a GitHub directory: https://github.com/CarlBethuel/Dempster-Shafer-work.International audienceThe remote sensing community benefits from new sensors and easier access to Earth Observation data to frequently released new land-cover maps. The propagation of such independent and heterogeneous products offers promising perspectives for various scientific domains and for the implementation and monitoring of land-use policies. Yet, it may also confuse the end-users when it comes to identifying the most appropriate product to address their requirements. Data fusion methods can help to combine competing and/or complementary maps in order to capitalize on their strengths while overcoming their limitations. We assessed the potential of the Dempster-Shafer Theory (DST) to enhance oil palm mapping in Sumatra (Indonesia) by combining four land-cover maps, hereafter named DESCALS, IIASA, XU, and MAPBIOMAS, according to the first author's name or the research group that published it. The application of DST relied on four steps: (1) a discernment framework, (2) the assignment of mass functions, (3) the DST fusion rule, and (4) the DST decision rule. Our results showed that the DST decision map achieved significantly higher accuracy (Kappa = 0.78) than the most accurate input product (Kappa = 0.724). The best result was reached by considering the probabilities of pixels to belong to the OP class associated with DESCALS map. In addition, the belief (i.e., confidence) and conflict (i.e., uncertainty) maps produced by DST evidenced that industrial plantations were detected with higher confidence than smallholder plantations. Consequently, Kappa values computed locally were lower in areas dominated by smallholder plantations. Combining land-use products with DST contributes to producing state-of-the-art maps and continuous information for enhanced land-cover analysis