Portail HAL des publications du LIRMM
Not a member yet
13279 research outputs found
Sort by
Applying the maximum entropy principle to neural networks enhances multi-species distribution models
No full textSubmitted to Methods in Ecology and EvolutionThe rapid expansion of citizen science initiatives has led to a significant growth of biodiversitydatabases, and particularly presence-only (PO) observations. PO data are invaluable for understanding species distributions and their dynamics, but their use in a Species Distribution Model (SDM) is curtailed by sampling biases and the lack of information on absences. Poisson point processes are widely used for SDMs, with Maxent being one of the most popular methods. Maxent maximises the entropy of a probability distribution across sites as a function of predefined transformations of variables, called features. In contrast, neural networks and deep learning have emerged as a promising technique for automatic feature extraction from complex input variables. Arbitrarily complex transformations of input variables can be learned from the data efficiently through backpropagation and stochastic gradient descent (SGD). Yet, deep learning was mainly developed for classification problems, and learning robust features and species abundances across space while properly correcting for sampling biases has remained a challenge so far. In this paper, we propose DeepMaxent, which harnesses neural networks to automatically learn shared features among species, using the maximum entropy principle. To do so, it employs a normalised Poisson loss where for each species, presence probabilities across sites are modelled by a neural network. We evaluate DeepMaxent on a benchmark dataset known for its spatial sampling biases, using PO data for calibration and presence-absence (PA) data for validation across six regions with different biological groups and covariates. Our results indicate that DeepMaxent performs better than Maxent and other leading SDMs across all regions and taxonomic groups. The method performs particularly well in regions of uneven sampling, demonstrating substantial potential to increase SDM performances. The method opens the possibility to learn more robust features predicting simultaneously many species to arbitrary large datasets without increased memory requirements. The model likelihood, arising from a Poisson process, makes the method compatible with the integration of more standardised types of data to further increase sampling bias correction. In particular, our approach yields more accurate predictions than traditional single-species models, which opens up new possibilities for methodological enhancement
T-CoLoc: Leveraging Tethers for Reliable Co-Localization within an Underwater ROV Chain
No full textInternational audienceUnderwater Remotely Operated Vehicles (ROVs) exchange data with a control station via a communication cable. One or more intermediate robots can be placed along this tether to manage its shape and minimize the mechanical effects on the ROV. This work deals with the localization of a pair of underwater robots connected by a tether, in a previously unknown environment. While each robot can estimate its trajectory and a model of its surroundings using Simultaneous Localization And Mapping (SLAM) algorithms, aligning these observations in the same reference frame requires inter-robot data association. In this work, we introduce T-Coloc, a new method for aligning models' frames that leverages an estimation of the tether shape to align individual robot observations. An experimental validation in a pool demonstrates that T-CoLoc can align the trajectories of the two robots in the same reference frame with an error lower than 20 cm using the noisy shape estimation of a 3 m long tether
RV-Sec5: Enhancing RISC-V Security Evaluation via Targeted ISA-Level Instrumentation using gem5
No full textInternational audienceThe modularity of the RISC-V Instruction Set Architecture (ISA) has accelerated its adoption in security-critical domains, yet it introduces significant challenges for pre-silicon security validation. Current evaluation methods often rely on high-level emulation that overlooks microarchitectural side effects or post-silicon testing that identifies vulnerabilities too late in the design cycle. This paper presents RV-Sec5, a systematic framework for ISA-level security evaluation that leverages the gem5 simulator. Unlike standard simulators, RV-Sec5 introduces a methodology to map high-level security invariants-such as privilege isolation and memory protection-directly to automated, cycle-accurate instrumentation points within the ISA decoder. This approach bridges the semantic gap between abstract security policies and low-level hardware execution. We demonstrate the framework's efficacy through a case study involving unauthorized Control and Status Register (CSR) modifications, showing how RV-Sec5 detects privilege escalation attempts and monitors microarchitectural anomalies, such as TLB flushes and cache state changes, in real-time.</div
Spectral Identification of Inertial Parameters in Forced Sinusoidal Regimes
No full textInternational audienceThis paper presents an approach for identifying all the inertial parameters of a solid (mass, position of the center of gravity, and inertia matrix) without repositioning the solid. The identification is based on the use of a hexapod parallel robot capable of six degree-of-freedom (DOF) motions and a 6–component force/torque sensor. The solid to be characterized is placed on the sensor, which is attached to the robot. The robot is used to impose different sinusoidal excitation trajectories in succession. The reaction forces are recorded at the same time, and the inertial parameters are identified by solving the Newton-Euler equations in the frequency domain using the Fourier transform. This solution in the frequency domain allows for precise computation with the following advantages: (i) maximum decoupling of the equations for optimal resolution, individually adapted to each parameter; (ii) simplicity of the experimental setup and the resolution method. Only three elements are required (the solid to be evaluated, the force/torque sensor, and the robot). Data processing is not overly complex and does not require overly restrictive synchronization of the clocks of the different systems; (iii) fine adjustment of the force/torque sensor (offset calibration) is not necessary
Formal Modeling and Analysis of Small-scale Data Centers Integrating Renewable Energy using Timed Automata
No full textInternational audienceIntegrating renewable energy into data centers is essential for reducing reliance on fossil fuels and minimize the environmental impact of digital infrastructures. However, the variability and unpredictability of renewable sources come with significant design and operational challenges. This paper introduces a formal modeling framework for solar-powered small-scale data centers, using stochastic timed automata and statistical model checking for mathematical analysis. The solution supports efficient resource sizing, reduces grid energy consumption through optimized workload scheduling and server renewal strategies. It enables robustness evaluation under component failure scenarios. A case study demonstrates the applicability, flexibility, and scalability of the framework for distributed system topologies and energy-aware design exploration
Post‐Release Survival of the Pelagic Stingray (Pteroplatytrygon violacea, Bonaparte, 1832) in French Longline Fisheries in the Northwestern Mediterranean Sea
No full textInternational audienceBycatch remains a critical challenge in global fisheries, even when using selective gears such as longlines. In the French longline fishery targeting Atlantic bluefin tuna ( Thunnus thynnus ) in the Gulf of Lion, the common pelagic stingray ( Pteroplatytrygon violacea ) is the primary bycatch species. This study investigated the post‐release survival and behaviour of 38 stingrays (38–75 cm disc width) captured during the spring–summer seasons of 2022 and 2023, using electronic tagging (MRPats, sPats, and PSATLife). A clear seasonal trend was observed, with smaller individuals more frequently caught in summer, likely linked to warmer water conditions that also reduced tag retention time (1–70 days). Survival was estimated using the Kaplan–Meier method, accounting for uncertainty in post‐release status determination. The results indicated high survival rates ranging from 73% to 100% (median 87%), demonstrating the species' strong resilience to capture and handling. Tagging data also revealed extensive vertical and horizontal movements, with individuals reaching depths of nearly 700 m and traveling over 20 km per day. This brings new and valuable information on this poorly known species, albeit common in the Mediterranean, for the sustainable management of exploited resources in this area
Environmental DNA data provides taxa-specific insights into community differences across Reunion Island back-reef depressions
No full textInternational audienceReunion Island’s coral reefs host complex biological communities, making comprehensive monitoring essential for detecting shifts in species presence as markers of reef health. Biodiversity monitoring, traditionally based upon visual censuses, can be enhanced through environmental DNA (eDNA) surveys that expand species detection. A recent eDNA metabarcoding study used 12S, 18S, and COI gene markers to screen water samples from four back-reef depressions (or ‘lagoons’) in Reunion Island, revealing distinctive communities across five different sites. This study offers new insight by identifying species and families driving community distinctiveness, and providing a preliminary glimpse into how single time-point eDNA data align with multi-year observational surveys from the Global Coral Reef Monitoring Network (GCRMN) in evaluating its effectiveness for detecting local fish species. The 12S marker detected 52 fish families across all sites, while COI and 18S, targeting mainly invertebrates, identified 262 and 414 families, respectively. Further analyses revealed strong contrasts among all sites, including two within the same lagoon. At the Planch’Alizé site, analyses indicated lower sequence read representation of corals, hydrozoans and several coral-associated fish species. Spatially proximate GCRMN surveys, aggregated over several years, recorded 149 fish species, compared to 213 detected by eDNA. Further, the GCRMN dataset recorded higher proportions of herbivores and omnivores, while eDNA detected a higher proportion of piscivores and captured more nocturnal and/or cryptic species. These results highlight the strength of eDNA for revealing fine-scale community differences, while emphasizing the importance of combining different survey types to gain a more comprehensive understanding of reef systems for supporting long term monitoring efforts
Impact of wheat-legume mix intercrops on wheat epidemics by modelling
No full textInternational audienceHighlights: • Simulated intercropping decrease disease intensity and improve protectiveness while canopy indicators predict such effects. • Pea intercropped with wheat decreased disease intensity compared with faba bean. • Nitrogen fertilization increased disease intensity. • This study stressed the critical lack of experimental data on disease in intercropping.Abstract: Context : Intercropping is a promising strategy for integrated disease management and agroecological transition, although experimental and modelling studies are scarce.Objectives: This study aims to understand and quantify the impact of non-host species choice and nitrogen (N) fertilization on disease epidemics in the context of intercropping.Methods: We collected existing experimental data on LAI based on a literature survey of non-diseased wheat intercropped with different non-host legume species (pea and faba bean) and N fertilization treatments. Based on a foliar epidemic model for intercropping, we simulated epidemics directly on these experimental data of LAI. The model is parameterized for two wheat fungal diseases: Septoria tritici blotch, a rain-borne disease, and wheat leaf rust, an air-borne disease.Results: Our results indicate that intercropping can decrease disease intensity and improve protectiveness for both diseases. Effect depends however on species choice as pea intercropped with wheat leads to lower disease intensity and better intercropping protectiveness compared with faba bean, whereas N fertilization increased disease intensity. We also found that crop indicators describing wheat leaf area index (LAI) can predict disease intensity, whereas indicators describing companion LAI can better predict intercropping protectiveness.Conclusions: Intercropping can significantly reduce fungal epidemics on wheat, and intercropping management practices can be optimized for effective disease management in wheat-legume intercrops. The dilution effect is more related to disease intensity, while the barrier effect is more related to intercropping protectiveness.Implications: These findings pave the way for identifying field indicators to predict epidemics. However, this study also stressed the critical lack of experimental data on disease in intercropping
Saturation-Based Adaptive Tracking Control of Underwater Vehicles: From Theoretical Design to Real-Time Experiments
No full textInternational audienceTracking control of an autonomous tethered underwater vehicle (ATUV) for a successful marine operation is a challenging task due to the complex and nonlinear dynamics of the vehicle characterized by parametric uncertainties. Besides these issues, the vehicle mainly operates in an uncertain and unpredictable environment. To deal with the ATUV control tracking problem, this article proposes a new tracking control approach that will be named saturation-based adaptive computed torque+ (SACT+). The proposed SACT+ is designed using a variable saturation function, a computed torque structure, a saturation-based dynamic feedback, and an adaptive mechanism. Then, several arguments, based on the well-known Lyapunov techniques, are proposed to prove the stability behavior of the final closed-loop dynamics. This ensures the convergence (theoretically) of the vehicle tracking error to the origin, leading to stable and safe operations. However, this tracking error (experimentally) only stays around the origin due to many factors, such as the measurement noise from the vehicle’s sensors, the inherent uncertainties of the vehicle combined with external disturbances from the marine environment, etc. Different tests are conducted in real-time using our underwater vehicle Leonard prototype to validate the proposed SACT+. The obtained experimental results show the effectiveness and robustness of the proposed SACT+ approach in real-life cases. Finally, the performance and energy consumption indices, as well as comparative experimental studies with two well-established controllers (from the literature), confirm the relevance of the proposed approach for controlling small-sized and/or low-cost underwater vehicles
Annotating Texture and Imitation Patterns in a Corpus of Slow Movements in Corelli’s Trio Sonatas
No full textInternational audienceThe Corelli trio sonatas are emblematic of the mid-Baroque period, characterized by their elegant melodies, intricate counterpoint, and systematic development of thematic material. Extending harmonic annotations from Hentschel et al. (2021), we provide a dataset with annotations of all 38 slow movements of the Sonate da chiesa Op. 1 (1681) and Op. 3 (1689), focusing on this Baroque writing style. The 7,000+ annotations describe the texture (homorhythmy, imitations, suspensions), further detailing thematic and imitation patterns, with other information such as rhythmic density. We provide audio synchronizations to a widely available recording and release the corpus for download as well as through the Dezrann platform