HAL-MINES ParisTech
Not a member yet
27348 research outputs found
Sort by
NERAF: 3D SCENE INFUSED NEURAL RADIANCE AND ACOUSTIC FIELDS
No full textInternational audienceSound plays a major role in human perception. Along with vision, it provides essential information for understanding our surroundings. Despite advances in neural implicit representations, learning acoustics that align with visual scenes remains a challenge. We propose NeRAF, a method that jointly learns radiance and acoustic fields. NeRAF synthesizes both novel views and spatialized room impulse responses (RIR) at new positions by conditioning the acoustic field on 3D scene geometric and appearance priors from the radiance field. The generated RIR can be applied to auralize any audio signal. Each modality can be rendered independently and at spatially distinct positions, offering greater versatility. We demonstrate that NeRAF generates high-quality audio on SoundSpaces and RAF datasets, achieving significant performance improvements over prior methods while being more dataefficient. Additionally, NeRAF enhances novel view synthesis of complex scenes trained with sparse data through cross-modal learning. NeRAF is designed as a Nerfstudio module, providing convenient access to realistic audio-visual generation.</div
Assessing Random Forest self-reproducibility for optimal short biomarker signature discovery
No full textInternational audienceBiomarker signature discovery remains the main path to develop clinical diagnostic tools when the biological knowledge on a pathology is weak. Shortest signatures are often preferred to reduce the cost of the diagnostic. The ability to find the best and shortest signature relies on the robustness of the models that can be built on such set of molecules. The classification algorithm that will be used is selected based on the average performance of its models, often expressed via the average AUC. However, it is not garanteed that an algorithm with a large AUC distribution will keep a stable performance when facing data. Here, we propose two AUC-derived hyper-stability scores, the HRS and the HSS, as complementary metrics to the average AUC, that should bring confidence in the choice for the best classification algorithm. To emphasize the importance of these scores, we compared 15 different Random Forests implementation. Additionally, the modelization time of each implementation was computed to further help deciding the best strategy. Our findings show that the Random Forest implementation should be chosen according to the data at hand and the classification question being evaluated. No Random Forest implementation can be used universally for any classification and on any dataset. Each of them should be tested for both their average AUC performance and AUC-derived stability, prior to analysis. Author summary To better measure the performance of a Machine Learning (ML) implementation, we introduce a new metric, the AUC hyper-stability, to be used in parallel with the average AUC. This AUC hyper-stability is able to discriminate ML implementations that show the same AUC performance. This metric can therefore help researchers in choosing the best ML method to get stable short predictive biomarker signatures. More specifically, we advocate a tradeoff between the average AUC performance, the hyper-stability scores, and the modeling time
Mobility, Mismatch, and Wages: An Analysis of Educational Returns
No full textInternational audienceMobility, Mismatch, and Wages: An Analysis of Educational Returns
Les start-ups « deep tech » : une réponse aux enjeux du XXIᵉ siècle ?
No full textThe ConversationDepuis quelques années, la « deep tech » a le vent en poupe. Mais que désigne ce terme ? Au-delà de la mode, le concept mobilise de vrais enjeux scientifiques, financiers mais aussi sociétaux
Oyster shells as archives of present and past environmental variability and life history traits: A multi‐disciplinary review of sclerochronology methods and applications
No full textInternational audienceOysters inhabit a variety of coastal and deep‐sea settings over a wide latitudinal range and have a role as ecosystem engineers. They also represent an important food source for humans since hunter‐gatherer times, which motivates interest in using oyster shells as environmental and life history archives. Still, oysters have often been disregarded in sclerochronology studies, although several methods based on both microtextural and geochemical approaches have successfully been investigated. We review how these methods have been used to improve interpretations of shell records, and we identify knowledge gaps in a variety of disciplines. Those include ecology to study larval dispersal and growth rates; archaeology to determine shell midden constructions and site occupations; and palaeoenvironmental and palaeoclimate reconstructions from tidal to annual timescales. We also suggest standardizing sclerochronology procedures to improve palaeoenvironmental reconstructions and biophysical models on oyster larval dispersal
Organisations et méthodes pour une rénovation frugale du patrimoine ferroviaire : une approche toposique de la validation dans les systèmes complexes
No full textThe renovation of aging infrastructures is crucial in the face of today's ecological and societal challenges. Designed for a defined lifespan with pre-established maintenance, these infrastructures will reach their operational limit in the coming decades. As their ageing leads to increased renovation costs, their future raises many questions in the face of budgetary constraints. In the rail sector, particularly for local rail lines, the dilemma is whether to accept their deterioration or invest in their renovation. The management model for these infrastructures, which separates investment from maintenance, is showing its limitations in the face of this patrimonial turn.This thesis, carried out at SNCF Réseau, aims to explore a rationalization of renovation that is underway and that would move towards a design-driven, innovative, frugal and sustainable renovation. This thesis shows that frugality is achieved through the unexpected activity of designing constructive proofs. It describes the organization set up to design such proofs, and establishes that it enables preserving innovation with propagation management for validation. Finally, this thesis models this renovation design using a toposic approach, in order to support the design of constructive proofs. Thus, this thesis reveals a new model of action, a creative preservation regime for frugal renovation, which allows us to think about the existence of objects without separating design and maintenance. This design regime cultivates a highly paradoxical form of knowledge renewal, utilization of existing knowledge and preservative innovation.La rénovation des infrastructures vieillissantes est cruciale face aux défis écologiques et sociétaux contemporains. Conçues pour une durée de vie définie avec une maintenance préétablie, ces infrastructures atteindront leur limite opérationnelle dans les prochaines décennies. Leur vieillissement entraînant une augmentation des coûts de rénovation, face aux contraintes budgétaires, leur futur soulève de nombreuses interrogations. Dans le secteur ferroviaire, notamment pour les lignes de desserte fine du territoire, se pose le dilemme de savoir s'il faut accepter leur dégradation ou investir dans leur rénovation. Le modèle de gestion de ces infrastructures, qui sépare les logiques d'investissement et celles de maintenance, montre ses limites face à ce tournant patrimonial.Dès lors, cette thèse, menée au sein de SNCF Réseau, vise à explorer une rationalisation de la rénovation qui serait en cours et qui irait vers une rénovation conceptive, innovante, frugale et durable. Cette thèse montre que la frugalité est obtenue au travers d'une activité inattendue, de conception de preuves constructives. Elle décrit l'organisation mise en place pour la conception de telles preuves, et établit qu'elle permet une innovation préservatrice avec une gestion de la propagation pour la validation. Enfin, cette thèse modélise cette conception de la rénovation en s'appuyant sur une approche toposique, afin d'accompagner la conception de preuves constructives. Ainsi, cette thèse révèle un nouveau modèle d'action, un régime de création préservatrice pour une rénovation frugale, qui permet de penser l'existence des objets sans séparer conception et maintenance. Ce régime de conception cultive une forme de renouvellement des savoirs, d'utilisation des savoirs existants et d'innovation préservatrice très paradoxale
: ‘Breaking the Mirror’ to Face Digital Convergence: The Role of Selective Mirroring in the Trade-Off between Value Creation and Capture Mechanisms
No full textThis research questions the mirroring hypothesis in the context of digital convergence. The mirroring hypothesis suggests that the organization of innovation activities tends to mirror the technical architecture of products. When the architecture is modular, such mirroring optimizes the management of innovation activities. But it can also limit the ability of incumbent firms to adapt to technological discontinuities. Digital convergence is a source of discontinuities that transform the conditions of value creation and capture within industries. It leads to new complementarities that push incumbents to collaborate with firms coming from other industries within emerging ecosystems. How does the mirroring between product architecture and organization evolve in the face of the new challenges of value creation and capture brought by digital convergence? This question is addressed through a qualitative case study of the organization of innovation activities between Renault and its partners in the field of embedded automotive electronics. The results show that the automaker ‘breaks the mirror’ through a strategy of selective mirroring that allows it to collaborate with new complementors and to reconfigure its mechanisms of value creation and capture.Xerfi Canal a reçu Alexandre Azoulay, Groupe de recherche en droit, économie et gestion (Gredeg), Université Côte d’Azur, CNRS, Sophia Antipolis, France, pour parler de l'hypothèse du miroir en management.Une interview menée par Jean-Philippe Denis
Advancing C3S Energy Services in support to Europe’s energy transition: co-development of the Pan-European Climate Database with ENTSO-E.
No full textInternational audienceThe Copernicus Climate Change Service (C3S), which is fully funded by the European Union and implemented by the European Centre for Medium-Range Weather Forecasts (ECWMF), has traditionally provided authoritative information about the past, present, and future climate through datasets, tools, and applications. The service facilitates the development of adaptation and mitigation strategies for society in the face of climate change.C3S has been consolidated as one of the most relevant providers of climate data and information both in Europe and globally. Its implementation has evolved contemporaneously with the evolution of the climate services arena. C3S has historically prioritized user requirements in shaping sectoral datasets and information delivered by the service. In recent years, however, it has taken additional measures to advance this commitment. These efforts include the development of customized climate datasets in partnership with core users. These users utilize and amplify the climate information supplied by C3S and disseminate it further through their own organizations and partners. This initiative involves prominent European institutions, such as the European Network of Transmission System Operators for Electricity (ENTSO-E), which plays an outstanding role in operating and securing the European electricity system and market integrations.In the context of Europe's energy transition towards a more environmental-friendly electricity supply system, the power sector is increasingly vulnerable to weather and climate variability. This heightened exposure underscores the need for collaborative efforts among climate and energy experts to develop information at the European level. In this context, the present contribution aims to illustrate the process of co-development undertaken by C3S and ENTSO-E for the preparation of the latest Pan-European Climate Database (PECD) version, that encompass both climate and energy data, considering for the first time, future climate projections. PECD is a dataset that has underpinned most studies conducted by TSOs ever since and plays an important role for the resilience of energy systems and optimization of their performance in response to climate change. In this contribution we will delve into the TSO's needs, engagement, and co-development process up to the final PECDv4.2 dataset, which is readily available at the Copernicus Climate Data Store, as well as the evolution towards the new version. This co-development has combined both academic and grid operators' needs and resulted in a dataset that is fit for research and industry alike. The contribution will highlight the lessons learned in the process, the features accomplished, and the remaining gaps
Travailler avec l’intelligence artificielle dans les industries culturelles et créatives.: Comment appliquer l'AI Act ?
No full textInternational audienceThe rise of Artificial Intelligence (AI) marks a major turning point in the transformation of industries and society, generating both significant opportunities and complex challenges. Its potential for automation and optimization is attracting businesses, including those in the Cultural and Creative Industries (CCI). However, concerns persist regarding its impact on traditional production methods, and it raises legal, ethical, and philosophical questions—particularly concerning the ownership of input and output data, as well as the role of human ingenuity in the industrial value chain. In response to these issues, the European Union has introduced the Artificial Intelligence Act (AI Act), which imposes transparency and safety requirements on companies. This volume synthesizes the discussions held at the symposium “How Can Industries Operate Under the AI Act?”, organized in Paris by the Robotics Center of Mines Paris – PSL on June 13–14, 2024. It explores the AI Act’s impact on the industrial sector, the future of manual labor, and the implications for cultural and creative industries. Through diverse testimonies, this manifesto opens a critical dialogue between researchers, legal experts, artisans, and policymakers, aiming to explore ways to reconcile AI-driven innovation with the traditions of cultural and creative industries for the benefit of society.L'essor de l'intelligence artificielle (IA) constitue un tournant majeur dans la transformation des industries et de la société, engendrant à la fois des opportunités significatives et des défis complexes. Son potentiel d'automatisation et d'optimisation attire les entreprises, y compris celles des industries culturelles et créatives (ICC). Cependant, des préoccupations subsistent quant à son impact sur les méthodes de production traditionnelles et elle soulève des interrogations légales, éthiques et philosophiques, notamment en ce qui concerne la propriété des données d'entrée et de sortie, ainsi que le rôle de l'ingéniosité humaine dans la chaîne de valeur industrielle. Pour répondre à ces enjeux, l'Union européenne a introduit l'Artificial Intelligence ACT (AI ACT), qui impose des exigences de transparence et de sécurité aux entreprises. Cet ouvrage synthétise les discussions du Symposium « Comment les industries peuvent-elles opérer à travers l'AI Act ? », organisé à Paris par le Centre de Robotique de Mines Paris – PSL, les 13 et 14 juin 2024. Il aborde l'impact de l'AI Act sur le secteur industriel, l'avenir des emplois manuels et les répercussions sur les industries culturelles et créatives. À travers des témoignages variés, ce manifeste ouvre un dialogue critique entre chercheurs, juristes, artisans et décideurs, visant à explorer les manières de concilier les innovations de l'IA avec les traditions des industries culturelles et créatives pour le bénéfice de la société