Portail HAL des publications du LIRMM
Not a member yet
13279 research outputs found
Sort by
In vivo deletion of a GWAS-identified Myb distal enhancer acts on Myb expression, globin switching, and clinical erythroid parameters in β-thalassemia
No full textInternational audienceGenome-wide association studies (GWAS) have identified numerous genetic variants linked to human diseases, mostly located in non-coding regions of the genome, particularly in putative enhancers. However, functional assessment of the non-coding GWAS variants has progressed at slow pace, since the functions of the vast majority of genomic enhancers have not been defined, impeding interpretation of disease-susceptibility variants. The HBS1L-MYB intergenic region harbors multiple SNPs associated with clinical erythroid parameters, including fetal hemoglobin levels, a feature impacting disease severity of beta-hemoglobinopathies such as sickle cell anemia and betathalassemia. HBS1L-MYB variants cluster in the vicinity of several MYB enhancers, altering MYB expression and globin switching. We and others have highlighted the conserved human MYB − 84kb enhancer, known as the − 81kb enhancer in the mouse, as likely candidate linked to these traits. We report here the generation of a Myb − 81kb enhancer knock-out mouse model, and shed light for the f irst time on its impact on steady state erythropoiesis and in beta-thalassemia in vivo
Comparative Study of Advanced Control Strategies of Cable-Driven Parallel Robots
No full textNational audienceCable-Driven Parallel Robots (CDPRs) offer large workspaces and high payload capacities, but their control remains challenging due to cable elasticity, unidirectional actuation, and model uncertainties.This paper presents a comparative study of three control strategies—Computed Torque Control (CTC), linear Model Predictive Control (MPC), and Sliding Mode Control (SMC)—applied to a 6-DoF CDPR with elastic cable. Each method is evaluated in simulation on the HRPcable CDPR, considering track-ing accuracy, robustness, computational cost, and implementation complexity. Particular attention is given to the integration of tension distribution and the controllers ability to handle singularities and tension feasibility constraints. Results highlight trade-offs between precision, robustness, and real-time feasibility, offering guidance for controller selection in practical CDPR applications
Biosealight: A Low-Cost, Compact, Autonomous, Low-Energy, High-Sensitivity Sensor for Detecting Marine Bioluminescence
No full textInternational audienceBioluminescence, the light emitted naturally by marine organisms, is an important way of communication between organisms in the mesopelagic zone. Nearly 75% of marine organisms, from the surface to the deep sea, and accross a wide taxonomic diversity, use this capability for communication with diverse ecological goals. Bioluminescence detection thus offers an indirect way of tracking the presence of organisms (ranging from zooplankton, dinoflagellates to fishes), their distribution and diel migrations. Such detection can lead, for example, to a better understanding of auto-and heterotrophic plankton communities, vertical migrations of organisms and consequently of a better quantification of the active carbon export in the mesopelagic ocean. However, current technologies still limit large deployments, and high acquisition frequency observations of in situ bioluminescence.This publication aims to describe a new low cost, compact, autonomous, and multi-sensors instrument developed to measure in situ bioluminescence: Biosealight.The sensor's acquisition system is built around a highsensitivity photon detector that detects photons and converts their number into proportional electrical signals. The photon detection system is integrated with additional sensors that record environmental, including pressure (rated up to 600 bar), temperature, and various system voltages. These measurements are managed by a motherboard featuring an STM32 microcontroller.By proposing another way to measure the bioluminescence in situ, and being easy to deploy, this instrument will increase the capacity to observe and charaterize bioluminescence of organisms and improve our understanding of its ecological role.</p
Lossless and Universal 3D Object Encryption With Differentiated Visual Effects Upon Decryption: A Novel Paradigm
No full textInternational audienceA recently proposed 3D object encryption scheme enables hierarchical decryption, allowing a single encryption to display varied visual effects upon decryption. This has potential applications in complex scenarios where 3D objects require access at different security levels. However, it has two main issues: lossless decryption is not possible due to precision loss from the IEEE 754 standard, and it is customized for a specific AES algorithm, limiting support for others. Motivated by this, a novel paradigm for 3D object encryption with differentiated visual effects upon decryption is proposed. In this paradigm, the precision loss is preserved within the 3D encrypted object, allowing the 3D object after fully decrypting to be identical to the original, and thus it is lossless. Meanwhile, the data to be encrypted is reduced to three blocks, which are generalized bitstreams. Bitstream encryption algorithms that do not produce ciphertext expansion can be applied, and it operates independently of the other components in the paradigm, making it universal. Finally, a prototype of this paradigm is constructed, using the same encryption algorithm as in the previous scheme for comparative experiments. Two 3D object datasets are used to conduct the experiments and results demonstrate that it achieves lossless decryption and has a time advantage, while also undergoing security analysis and verification against smoothing attacks to confirm the security
MorphoNet 2.0: An innovative approach for qualitative assessment and segmentation curation of large-scale 3D time-lapse imaging datasets
No full textInternational audienceThanks to recent promising advances in AI, automated segmentation of imaging datasets has made significant strides. However, the evaluation and curation of 3D and 3D+t datasets remain extremely challenging and highly resource-intensive. We present MorphoNet 2.0, a major conceptual and technical evolution designed to facilitate the segmentation, self-evaluation, and correction of 3D images. The application is accessible to non-programming biologists through user-friendly graphical interfaces and works on all major operating systems. We showcase its power in enhancing segmentation accuracy and boosting interpretability across five previously published segmented datasets. This new approach is crucial for producing ground-truth datasets of discovery-level scientific quality, critical for training and benchmarking advanced AI-driven segmentation tools, as well as for competitive challenges
Empowering Relational Concept Analysis Using Large Language Model Knowledge Delivery
No full textInternational audienceRelational concept analysis (RCA) computes and represents relational datasets as lattices of concepts and implications. This approach relies on the computation of relational attributes which capture quantified relationships between objects and concepts. However, interpreting this representation is challenging for end-users. FCA experts must combine relational attributes from different concept lattices to obtain a complete formulation, while domain experts (e.g., software engineers, agronomists) face the additional difficulty of extracting meaning from these mathematical structures. This paper proposes an approach to address this issue by leveraging Large Language Models (LLMs). It comprises two main components: (1) the generation of a textual description of relational attributes; (2) the design of a generic prompt that enables LLM to produce a natural-language description of RCA artifacts using the generated textual description. The approach is illustrated using representative implications from a real-world agroecological dataset
A Concise Ontological Model of the Design and Optoelectronic Properties in the Quantum Cascade Laser Domain
No full textInternational audienceTerahertz quantum cascade lasers (QCLs) are semiconductor laser devices that operate in the far infrared range (frequency range from about 100 GHz to 10 THz). QCL properties can be categorized as follows: design of the laser (heterostructure properties capturing the various materials used in the laser structure and the various laser design types) and the laser Optoelectronic properties (laser performance behavior as a result of injection of current into the laser device). Maintaining ontologies with this information is useful in supporting data mining activities that seek to retrieve useful information on the various QCL designs and their respective performance, together with provenance information. This provides a platform to share and interact with QCL data by both machines and humans in a Findable, Accessible, Interoperable, and Reusable manner. The existing ontologies in the material design domain do not capture this crucial information. This is due to a lack of formal definitions for the QCL property concepts. In this paper, we address the issue of formal representation of the specified QCL properties and the relationships among them. We propose a semantically enriched ontological model of properties in the QCL domain. We evaluate the ability of ontological representation to model the QCL properties using an inheritance richness metric-based evaluation and the ontology validation technique. Experimental evaluation indicates the consistency of the ontology, its ability to answer 100% of the competency questions by QCL domain experts, and an inheritance richness metric of 0.133, indicating a detailed level of the ontology in capturing the domain requirements
LLMs Do It All: A Reality Check with Formal Concepts
No full textOral communication at the workshop ConSOFT (Conceptual Knowledge Software: Recent Advancements and Examples), during the 2nd International Joint Conference on Conceptual Knowledge Structures (https://www.kde.cs.uni-kassel.de/consoft/)Following the rise and growing capabilities of LLMs, a friendly colleague recently questioned us about the relevance of continuing to write efficient algorithms and implementing them in frameworks, or even using formal concept analysis! We took him at his word, and since the topic is broad, we began by asking: can LLMs compute the full set of concepts of a formal context — a task known to be highly combinatorial? We explored this question using two approaches: the first ("direct computation") consisted of asking the LLMs directly for the list of concepts, while the second ("code generation") involved asking them to produce code that performs the computation, which we then tested. We selected a benchmark of formal contexts of various sizes, several LLM models, and multiple prompting strategies. In addition, to the non-negligible economic and environmental costs that must be considered, the results are not yet conclusive. Direct computation performs very poorly, with heavily degraded results for contexts with more than 15 objects and 15 attributes. Code generation yields better performance, but results remain imperfect, and computation times are significantly higher compared to the reference tool used for comparison (fca4j). While this doesn't predict the future, it offers a snapshot of the current state of progress, and the approach we implemented can be periodically re-evaluated. This study was conducted in collaboration with students, simulating the behavior of junior developers or non-expert users who may not have prior knowledge in Formal Concept Analysis or prompt engineering. Our objective was not only to evaluate the technical performance of LLMs, but also to explore their accessibility and usefulness in educational and low-expertise contexts. In this regard, the work aims to demonstrate whether LLMs can serve as a meaningful support tool even for those without strong computational backgrounds, thereby extending the benefits of FCA-based reasoning and concept structuring to a broader audience
All Kolmogorov complexity functions are optimal, but are some more optimal?
No full textKolmogorov (1965) defined the complexity of a string as the minimal length of a program generating . Obviously this definition depends on the choice of the programming language. Kolmogorov noted that there exist optimal programming languages that make the complexity function minimal up to additive terms, and we should take one of them — but which one? Is there a chance to agree on some specific programming language in this definition? Or at least should we add some other requirements to optimality? What can we achieve in this way?In this paper we discuss different suggestions of this type that appeared since 1965, specifically a stronger requirement of universality (and show that in many cases this does not change the set of complexity functions)
Improving Yolov8 For Fast Few-Shot Object Detection By Dinov2 Distillation
No full textInternational audienceRecent neural networks for object detection achieve excellent performance when trained on large databases but still struggle to learn new objects with few examples, leading to the development of Few-Shot Object Detection (FSOD). State-of-the-art FSOD methods often use computationally heavy architectures like Faster R-CNN or pre-trained Vision Transformers (ViTs) such as DINOV2, limiting their usage for real-time inference in resource-constrained environments. We propose a novel approach that transfers rich features from a pre-trained ViT (DINOV2) to the lightweight YOLOv8 architecture via knowledge distillation. This bridges the gap between FSOD performance and efficiency, enabling YOLOv8 to better handle few-shot scenarios while retaining computational efficiency. Experiments on the MSCOCO benchmark adapted for FSOD show that our method enhances the performance of lightweight detectors, highlighting the benefits of combining ViT feature learning with efficient detectors for real-world FSOD