Portail HAL des publications du LIRMM
Not a member yet
13279 research outputs found
Sort by
M4.5 - Internal and external use case evaluation and demonstrators
No full textThis report describes work under the FAIR-IMPACT project WP4/T4.5 FAIR semantic artefacts in use within data repositories, and its main focus points: to demonstrate the feasibility of setting up a connector (Semantic Artefacts Catalogues - SACs x (meta)data repository) and to provide concrete evidence of the added value provided by these SAs for the FAIRness of the scientific data stored in the repositories. A number of use cases (UCs) from different scientific disciplines (agri-food, life sciences, ecology, technical and physical sciences, archaeology, Social Sciences and Humanities, astronomy) and operating at different scales and in different research infrastructures (RI) were examined. They identified the semantic needs of their respective communities and the state of progress and organisation of the semantic representation of knowledge in their environments. Connectors have been (or will be) incorporated into (meta)data repositories depending on : Metadata that can be semantized SAs and SACs available in the community The technologies used by (meta)data repositories and by SAs and SACs The needs and practices of (meta)data repositories users: data users, depositors and manager
Romo : Romo is a software (research prototype) that aims to extract component-based architecture based on the analysis of the object-oriented JAVA source code.
No full textRomo : Romo is a software (research prototype) that aims to extract component-based architecture based on the analysis of the object-oriented JAVA source code
Hybrid AI with LLMs and Theorem Provers for Semantic Parsing and Natural Language Inference for French
No full textInternational audienc
Toward Trustworthy Automated Driving through Qualitative Scene Understanding and Explanations
No full textInternational audienceUnderstanding driving scenes and communicating automated vehicle decisions are key requirements for trustworthy automated driving. In this article, we introduce the qualitative explainable graph (QXG), which is a unified symbolic and qualitative representation for scene understanding in urban mobility. The QXG enables interpreting an automated vehicle’s environment using sensor data and machine learning models. It utilizes spatiotemporal graphs and qualitative constraints to extract scene semantics from raw sensor inputs, such as LiDAR and camera data, offering an interpretable scene model. A QXG can be incrementally constructed in real-time, making it a versatile tool for in-vehicle explanations across various sensor types. Our research showcases the potential of QXG, particularly in the context of automated driving, where it can rationalize decisions by linking the graph with observed actions. These explanations can serve diverse purposes, from informing passengers and alerting vulnerable road users to enabling post hoc analysis of prior behaviors
Cost-Effective Analytical Models of Resistive Opens Defects in FinFET Technology
No full textInternational audienceFinFET technology has become an attractive candidate for high-performance and power-efficient applications. However, its susceptibility to defects increases due to the complexity of the process fabrications and smaller feature sizes. This article proposes compact and low-cost analytical models to evaluate the delay increase in FinFET-based circuits due to resistive open defects. The models rely on electrical simulations to precharacterize the circuit library. Analytical expressions are developed for the three types of resistive opens that may occur in FinFET-based logic cells using multifin and multifinger structures. These types of resistive opens include: a resistive open at the drain or source of the transistors (RODS), a resistive open affecting the gate of a single transistor, and a resistive open affecting the gates of both nMOS and pMOS transistors. Compact analytical models are also developed to evaluate the delay increase due to the resistive open defects under process variations. Independent and correlated process variations are taken into account. The analytical models have been validated against SPICE electrical simulations. The proposed analytical models can be used to evaluate the detectability of resistive open defects, significantly reducing the cost of dealing with different defect sizes. Potential applications of the developed analytical models are delineated. This work allows us to have higher quality and reliable electronic products
Multi-objective Optimal Trajectory Generation for a Kite-based Ship Traction System
No full textInternational audienceThe use of kites for ship traction represents an innovative and promising solution to reduce greenhouse gas emissions in the maritime shipping sector, which is responsible for approximately 3% of global CO² emissions [3]. By leveraging wind energy, kite systems offer potential solution to lower fuel consumption [6] and operational costs. Existing approaches primarily focus on optimizing traction without explicitly considering energy consumption [4, 5]. While maximizing traction enhances propulsion efficiency most of the time, the energy required to control the kite, particularly through the winch motors, is a critical factor often overlooked. The present study addresses this lack by integrating energy consumption into the optimization process [2], this work is conducted as a part of Kiwin poject withBeyond the sea company (figure 1) and is funded by BPI france. The goal is then to enhance overall system performance while providing insights into the trade-off between traction and energy consumption. The proposed solution combines these two goals into a single optimization framework
Algorithmes pour l'énumération de mots b-canoniques
No full textNational audienceThe border array is the data structure built during the preprocessing of a word in the Morris-Pratt pattern matching algorithm. It is a key data structure for string algorithms. Several articles have proposed algorithms to recognize a border array, that is to determine if an array of integers is the border array of a word. In 1999, Moore et al. defined the notion of b-equivalence between words of length n as follows: two words are b-equivalent if they share the same border array. The relation of b-equivalence is an equivalence relationship. A word is said to be b-canonical if it is the smallest of its b-equivalence class in lexicographic order. We propose a linear space algorithm to enumerate all (distinct) b-canonical words of length n; this improves on the algorithm of Moore et al. We also present a parallel version of this algorithm and evaluate its performance compared to the non parallel version.This work improves on the algorithm proposed by Moore et al and also answers related combinatorial questions, such as the number of words belonging to the b-equivalence class of a b-canonical word. Enumerating b-canonical words and their border arrays for a given length yields an optimal test set of instances for any algorithm computing border arrays. This allows to create unit tests with complete coverage and zero redundancy
Transductive Conformal Inference for Full Ranking
No full textInternational audienceWe introduce a method based on Conformal Prediction (CP) to quantify the uncertainty of full ranking algorithms. We focus on a specific scenario where items are to be ranked by some ``black box'' algorithm. It is assumed that the relative (ground truth) ranking of of them is known. The objective is then to quantify the error made bythe algorithm on the ranks of the new items among the total . In such a setting, the true ranks of the original items in the total depend on the (unknown) true ranks of the new ones. Consequently, we have no direct access to a calibration set to apply a classical CP method. To address this challenge, we propose to construct distribution-free bounds of the unknown conformity scores using recent results on the distribution of conformal p-values. Using these scores upper bounds, we provide valid prediction sets for the rank of any item. We also control the false coverage proportion, a crucial quantity when dealing with multiple prediction sets. Finally, we empirically show on both synthetic and real data the efficiency of our CP method for state-of-the-art algorithms such as RankNet or LambdaMart
ADAM: ADAptive microcontroller platform for edge AI systems
No full textInternational audienceModern edge computing systems must balance performance and energy efficiency under constrained power budgets. We present ADAM (ADAptive Microcontroller), an open-source and parametric RISC-V platform designed to explore architectural and softwarelevel mechanisms for energy-aware embedded computing. ADAM supports heterogeneous cores organized into separate low-power and high-performance domains, enabling partitioned execution across a range of workload intensities. The platform introduces the Activity Pause Protocol, a software-driven interface for coordinating clock and power gating transitions, and includes a hybrid evaluation flow that combines FPGA execution with post-synthesis RTL simulation using execution context snapshotting. We demonstrate ADAM's capabilities with a keyword spotting use case, highlighting how domain separation and programmable power management reduce energy consumption by up to 38% in intermittent workloads
Utilisation de mécanismes inférentiels dans le processus d'explication automatique de la métaphore à une inconnue
No full textet 32e Conférence sur le Traitement Automatique des Langues Naturelles (TALN)International audienceWe consider metaphors to be analogies with one unknown. Explaining it means solving the variable of the resulting analogical square, the other three terms beeing fixed. We propose a detailed method to achieve this goal using the knowledge base JeuxDeMots. We proceed by recognizing previously identified relational patterns that allow us to evaluate the strength of the relational similarity and those of the two attributional similarities to deduce that of the analogy as a whole. The candidate term that allows us to obtain the best strength of analogy between the four terms of the analogy with the gap thus completed is elected. Finally, we seek to demonstrate that the use of inferences in this process allows us to achieve better results, that is, to increase the number of times a good candidate is elected.Considérons la métaphore comme une analogie à une inconnue. L'expliquer revient à résoudre l'unique variable du carré analogique qui en résulte et dont les trois autres termes sont fixés. Nous proposons ici une méthode détaillée pour arriver à cet objectif en utilisant la base de connaissances JeuxDeMots. Nous procédons par reconnaissance de schémas de relations préalablement identifiés et qui permettent d'évaluer la force de la similarité relationnelle et celles des deux similarités attributionnelles pour en déduire celle de l'analogie dans sa globalité. Le terme candidat qui permet d'obtenir la meilleure force d'analogie entre les quatre termes de l'analogie à trou ainsi complétée est élu. Enfin, on cherche à démontrer que l'utilisation d'inférences dans ce processus permet d'aboutir à de meilleurs résultats, c'est à dire augmenter le nombre de fois où un bon candidat est élu